1008-00005 · Lumos Analytics Inc · Strategy & Analysis

damages-analysis-memo.docx DOCUMENT Litigation

34 / 35 ← → ✕

PRIVILEGED AND CONFIDENTIAL — ATTORNEY WORK PRODUCT

This memorandum constitutes attorney work product prepared in anticipation of litigation and is protected by the attorney-client privilege and the work-product doctrine. It is intended solely for the use of the attorneys and client identified herein and should not be disclosed to any third party without the express consent of the undersigned.

CALDERWOOD & HARKNESS LLP INTERNAL MEMORANDUM

FROM: Wendy Schultz, Lead Associate, and Omar Siddiqui, Working Associate

TO: Marcus Delgado, Responsible Partner, and Catherine Bloom, Billing Partner

DATE: August 15, 2023

RE: Damages Analysis and Settlement Valuation — Lumos Analytics Inc. v. NovaSpark Technologies Inc. et al., Case No. 23-CV-004871 (Superior Court of California, County of Santa Clara)

MATTER NUMBER: 1008-00005

I. Executive Summary

The purpose of this memorandum is to present a comprehensive damages analysis for the trade-secret misappropriation claims asserted by our client, Lumos Analytics Inc. ("Lumos"), against NovaSpark Technologies Inc. ("NovaSpark"), Rajiv Anand, Serena Mallick, and Derek Yun (collectively, "Defendants"), and to recommend a settlement range and negotiation strategy in advance of the mediation session scheduled for September 6, 2023 before the Honorable Eleanor Voss (Ret.) at Tidewater Mediation Group.

Bottom Line. Based on our analysis and the technical input received from Lumos's retained consultants, the unjust enrichment conferred upon NovaSpark through the misappropriation of Lumos's proprietary technology is estimated at $8.7 million to $12.4 million, reflecting approximately 14 to 18 months of accelerated development time that NovaSpark obtained by using Lumos's trade secrets rather than developing competing technology independently. We recommend a target settlement range of $4 million to $6 million in monetary compensation, together with a stipulated permanent injunction and a comprehensive forensic remediation protocol to be supervised by Brindoval Digital Forensics LLC ("Brindoval"), with technical oversight from Naveen Kapoor of Meridian Cybersecurity Advisors Inc. ("Meridian"). These non-monetary terms are essential to achieving Lumos's primary business objective of protecting its proprietary technology from ongoing exploitation.

The damages analysis set forth in this memorandum incorporates technical input from Naveen Kapoor, Senior Consultant at Meridian, Lumos's retained cybersecurity and technical consultant, and builds upon the findings of the Brindoval Preliminary Forensic Report dated May 1, 2023. The analysis is further informed by deposition testimony obtained from each of the three individual defendants, financial disclosures produced by NovaSpark in discovery, and publicly available information regarding NovaSpark's funding and operations.

The claims at issue in this litigation are as follows:

1. Misappropriation of trade secrets under the California Uniform Trade Secrets Act ("CUTSA"), Cal. Civ. Code §§ 3426–3426.11;

2. Breach of contract — specifically, each individual defendant's Proprietary Information and Inventions Assignment Agreement ("PIIAA") executed upon commencement of employment with Lumos;

3. Breach of the duty of loyalty owed by each individual defendant to Lumos during the period of their employment;

4. Unfair business practices in violation of California Business and Professions Code § 17200 et seq.; and

5. Conversion of Lumos proprietary materials and intellectual property.

This memorandum proceeds in seven parts. Part II summarizes the key evidence developed through discovery to date. Part III presents the damages analysis, including the primary unjust enrichment theory and alternative damages theories. Part IV provides a settlement valuation that accounts for litigation risks, collection concerns, and cost-of-litigation considerations. Part V outlines our recommended mediation strategy. Part VI summarizes fees and costs incurred to date and projected through trial. Part VII sets forth our recommendation for next steps.

II. Summary of Key Evidence Developed to Date

The evidence gathered through forensic investigation, document discovery, and deposition testimony provides a strong foundation for Lumos's claims. Below, we summarize the principal evidentiary categories and their significance to the damages analysis and settlement posture.

A. Forensic Evidence — Brindoval Preliminary Forensic Report (May 1, 2023)

Brindoval Digital Forensics LLC, led by Dr. Irene Chow, Managing Director, and Kevin Barr, Senior Forensic Analyst, conducted a comprehensive forensic investigation pursuant to the Court's March 8, 2023 stipulation and order. On May 1, 2023, Brindoval issued its Preliminary Forensic Report, which confirmed the presence of Lumos proprietary materials on NovaSpark's corporate servers, each individual defendant's personal computing devices, and each individual defendant's personal cloud storage accounts. The report is detailed and well-documented, and we believe it will be highly persuasive to both the mediator and, if necessary, the trier of fact.

The specific findings by defendant are summarized as follows:

Rajiv Anand. Brindoval confirmed that on September 26, 2022 — four days before his resignation from Lumos — Anand transferred approximately 3.2 gigabytes of files from his Lumos-issued workstation to a personal Hollowell Health IT account. The transferred files included Lumos's proprietary model architecture specifications and hyperparameter configuration files for the LumosEdge ensemble machine-learning model, which constitutes the core of Lumos's commercial product offering. Additionally, on September 28, 2022, Anand forwarded a 37-page confidential document titled "LumosEdge 3.0 Product Roadmap" in PDF format to his personal email address at r.anand.personal@vaultmail.com. The Product Roadmap contained Lumos's strategic plans for the 2023–2025 product development cycle, including unreleased feature specifications and competitive positioning analysis. Anand did not have authorization to transfer any of these materials to personal accounts, and Lumos's information security policies expressly prohibit such transfers.

Serena Mallick. On October 10, 2022 — four days before her resignation from Lumos — Mallick cloned the entirety of the DataStream ETL pipeline repository, comprising approximately 780 megabytes of code, configuration files, and data-schema definitions, to a personal external USB drive bearing the serial number WD-4837291. The DataStream ETL pipeline is one of Lumos's most significant proprietary assets, as it manages the ingestion, transformation, and loading of data from 43 separate alternative-data vendor integrations. The pipeline took Lumos's engineering team approximately three years to develop and refine, and it represents a substantial competitive advantage in the alternative-data analytics market. Brindoval confirmed that the USB drive was subsequently connected to a computing device registered to NovaSpark's corporate network.

Derek Yun. On October 18, 2022 — two days before his resignation from Lumos — Yun downloaded 47 proprietary Jupyter notebooks from Lumos's internal development environment to his personal Larkspur Technologies account. These notebooks contained Lumos's proprietary feature-engineering code and model-training scripts, which together embody the accumulated work of Lumos's data science team over a period of approximately four years. The notebooks contain code for generating and validating 2,847 individually engineered features used in Lumos's financial signal-generation models. Brindoval's analysis of metadata and access logs confirmed that these notebooks were subsequently accessed and utilized within NovaSpark's development environment.

Across all three defendants, Brindoval's metadata and timestamp analysis corroborates the exfiltration timeline set forth above. The report demonstrates a coordinated pattern of data extraction occurring within a compressed timeframe — from September 26 through October 18, 2022 — immediately preceding each defendant's resignation from Lumos. The forensic evidence further confirms that the misappropriated materials were not merely downloaded and retained but were actively accessed and incorporated into NovaSpark's development workflow. This finding is critical because it establishes not only the act of misappropriation but also the actual use of Lumos's trade secrets in developing a competing product.

B. Deposition Testimony

We conducted depositions of each of the three individual defendants during May and June 2023. The testimony obtained is summarized below and includes several significant admissions that substantially strengthen Lumos's case.

Rajiv Anand (May 15, 2023). Anand acknowledged transferring files from his Lumos workstation to his personal Hollowell Health IT account on September 26, 2022, but characterized the transferred materials as "general reference materials" and "industry knowledge" that he believed he was entitled to retain. This characterization is directly contradicted by the forensic evidence, which confirms that the transferred files included Lumos's proprietary model architecture specifications and hyperparameter configurations — materials that are clearly identified as confidential in Lumos's information classification system and that bear proprietary markings. Anand further admitted that he coordinated his departure from Lumos with Mallick and Yun, and that the three of them had discussed forming a new venture together prior to resigning. Critically, corporate records produced in discovery confirm that Anand incorporated NovaSpark Technologies Inc. with the California Secretary of State on October 3, 2022 — just three days after his September 30, 2022 resignation from Lumos. This timeline strongly suggests premeditation and planning while still employed by Lumos.

Serena Mallick (June 7, 2023). Mallick admitted during her deposition that she copied the DataStream repository to the external USB drive (serial number WD-4837291) on October 10, 2022, but claimed that she did so merely to "retain personal notes" that she had created during her employment at Lumos. This explanation is not credible. The DataStream ETL pipeline repository consists of approximately 780 megabytes of complex production code, configuration files, and schema definitions covering data ingestion from 43 alternative-data vendor integrations. It is not a collection of personal notes. The repository represents a comprehensive, integrated codebase that would be of enormous value to any competitor seeking to rapidly establish alternative-data ingestion capabilities. Mallick further confirmed that she commenced employment at NovaSpark as Vice President of Engineering on October 17, 2022, just three days after her October 14, 2022 resignation from Lumos. She did not take any period of separation between the two positions.

Derek Yun (June 22, 2023). Yun's deposition yielded what we regard as the most significant evidentiary admission in the case. Yun acknowledged downloading 47 proprietary Jupyter notebooks from Lumos's internal systems to his personal Larkspur Technologies account on October 18, 2022. More importantly, and unlike his co-defendants, Yun went further and admitted that he subsequently used the downloaded notebooks in NovaSpark's product development. Specifically, Yun testified that the notebooks served as a "starting point" and "reference framework" for developing NovaSpark's competing feature-engineering capabilities. This admission is devastating to NovaSpark's independent-development defense and establishes actual use of Lumos trade secrets in the development of a competing product — a critical element under CUTSA. We expect Defendants' counsel will attempt to minimize the significance of this admission, but the deposition transcript is clear and unambiguous.

PIIAA Obligations. Each of the three individual defendants executed Lumos's standard Proprietary Information and Inventions Assignment Agreement upon commencement of their employment. The PIIAA contains the following material restrictive covenant provisions:

(a) Assignment of Inventions. All inventions, discoveries, developments, and improvements conceived or made by the employee during employment, or within twelve months thereafter if related to Lumos's business, are assigned to Lumos.

(b) Confidentiality and Non-Disclosure. The employee is obligated to maintain the confidentiality of all Lumos proprietary information and trade secrets, with this obligation surviving termination of employment for a period of three years.

(c) Non-Solicitation. The employee is prohibited from soliciting Lumos employees for a period of twelve months following separation from employment.

(d) Covenant Not to Use or Disclose Trade Secrets. The employee covenants not to use or disclose Lumos trade secrets at any time, without limitation, consistent with California law governing the protection of trade secrets.

None of the three individual defendants returned company materials or confirmed the deletion of Lumos proprietary information from their personal devices and accounts upon their departure from Lumos, as required by Section 7 of the PIIAA. This failure to comply with the return-of-materials obligation is itself a breach of contract and further evidences the intent to retain and use Lumos's proprietary materials at NovaSpark.

C. Trade Secrets at Issue — Five Categories

Through the course of this litigation, we have identified five categories of Lumos trade secrets that were misappropriated by Defendants. These categories have been specifically identified in our trade-secret disclosure statement filed with the Court on January 30, 2023, as supplemented on April 14, 2023, and they form the basis for our damages analysis:

1. LumosEdge Ensemble ML Model Architecture. This includes the proprietary neural-network topology, ensemble model configuration, and hyperparameter configurations that define the core algorithmic engine of Lumos's commercial product. The model architecture reflects years of iterative development, extensive experimentation, and significant investment in research and development. Lumos has invested approximately $14.2 million in R&D costs over the past four fiscal years to develop and refine this architecture.

2. Proprietary Feature-Engineering Libraries. These libraries contain 2,847 individually engineered features for financial signal generation, representing the accumulated expertise of Lumos's data science team. Each feature was developed, tested, and validated through a rigorous multi-stage process involving backtesting across historical market conditions. The feature libraries are among Lumos's most closely guarded assets, as they directly influence the predictive accuracy and commercial value of the LumosEdge product.

3. DataStream ETL Pipeline Codebase and Data-Schema Definitions. The DataStream ETL pipeline manages the ingestion, transformation, and loading of data from 43 alternative-data vendors. The pipeline is a complex, integrated system that took Lumos's engineering team approximately three years to develop. It includes custom connectors for each vendor's API, data-cleaning and normalization routines, schema validation logic, and error-handling protocols. The pipeline's vendor integrations are a significant barrier to entry in the alternative-data analytics market.

4. Curated and Labeled Training Dataset. Lumos has assembled and curated a training dataset of approximately 14.6 terabytes of historical financial and alternative data. This dataset has been labeled and structured for use in training and validating Lumos's machine-learning models. While the underlying raw data may be obtained from third-party sources, the curation, labeling, structuring, and quality-assurance processes applied to the dataset represent substantial independent value and constitute trade-secret-protectable work product.

5. Confidential Product Roadmap Documents (2023–2025). These documents include the 37-page "LumosEdge 3.0 Product Roadmap" PDF that Anand forwarded to his personal email, as well as related strategic planning documents. The roadmap contains unreleased product specifications, feature prioritization analysis, competitive intelligence, market positioning strategy, and planned vendor partnerships. Disclosure of this information to a competitor provides a significant strategic advantage by revealing Lumos's planned direction of development.

Each of these categories satisfies the statutory definition of a trade secret under CUTSA, Cal. Civ. Code § 3426.1(d), as information that derives independent economic value from not being generally known or readily ascertainable and that has been the subject of reasonable efforts to maintain its secrecy. Lumos maintains comprehensive information security measures, including role-based access controls, encryption, secure development environments, and the PIIAA obligations discussed above.

III. Damages Analysis

A. Applicable Legal Framework

The California Uniform Trade Secrets Act provides for multiple measures of damages for trade-secret misappropriation. Under Cal. Civ. Code § 3426.3, a plaintiff may recover damages measured by one or more of the following:

(i) Actual Loss. The actual loss to the plaintiff caused by the misappropriation, including both direct losses and consequential damages. Cal. Civ. Code § 3426.3(a).

(ii) Unjust Enrichment. The unjust enrichment caused by the misappropriation that is not taken into account in computing actual loss. This measure allows recovery of the benefit conferred upon the defendant through the wrongful use of the plaintiff's trade secrets, even where the plaintiff's own losses are difficult to quantify. Cal. Civ. Code § 3426.3(a).

(iii) Reasonable Royalty. In lieu of damages measured by any other method, the court may order payment of a reasonable royalty for no longer than the period of time the use could have been prohibited. Cal. Civ. Code § 3426.3(b). This measure is typically applied as a floor or alternative when actual loss and unjust enrichment are difficult to establish.

Additionally, Cal. Civ. Code § 3426.3(c) provides that if willful and malicious misappropriation is proven, the court may award exemplary damages in an amount not exceeding twice any compensatory damages award. This provision serves both a punitive and a deterrent function and is an important component of Lumos's overall damages exposure framework for purposes of settlement negotiations.

For purposes of this analysis, we focus primarily on the unjust enrichment theory as the most readily quantifiable and persuasive measure of damages given the evidence currently available. The unjust enrichment approach is well-suited to this case because: (a) the evidence clearly establishes that NovaSpark used Lumos's trade secrets to accelerate its product development, (b) the development-time savings can be quantified with reasonable certainty based on technical expert analysis, and (c) NovaSpark's financial records provide a market-based anchor for valuing the time saved. We discuss alternative theories — actual loss and reasonable royalty — below, as well as the exemplary damages multiplier.

B. Unjust Enrichment Analysis — Accelerated Time-to-Market

Methodology. The core of our damages analysis rests on quantifying the development time that NovaSpark saved by misappropriating and using Lumos's proprietary materials, rather than developing competing technology independently through its own research and engineering efforts. This "accelerated time-to-market" approach is well-established in trade-secret litigation and has been accepted by California courts as a basis for calculating unjust enrichment. See, e.g., Sargent Fletcher, Inc. v. Able Corp., 110 Cal. App. 4th 1658, 1670 (2003) (unjust enrichment may be measured by the "development costs the defendant avoided through misappropriation"); Ajaxo Inc. v. ETrade Group, Inc.*, 135 Cal. App. 4th 21, 57–58 (2005).

Our retained technical consultant, Naveen Kapoor, Senior Consultant at Meridian Cybersecurity Advisors Inc., conducted an independent assessment of the development timeline that NovaSpark would have required to replicate the functionality provided by the five categories of misappropriated Lumos trade secrets, assuming NovaSpark had proceeded entirely through independent development without access to any of Lumos's proprietary materials. Mr. Kapoor's assessment is based on his review of the Brindoval forensic findings, his independent analysis of the complexity of the misappropriated technologies, his experience in evaluating the development timelines for comparable machine-learning and data-engineering systems in the financial technology sector, and publicly available information regarding NovaSpark's team composition and technical capabilities.

Time Saved — Estimated at 14 to 18 Months. Based on Mr. Kapoor's technical assessment, the total development time saved by NovaSpark through its use of Lumos's proprietary materials is estimated at 14 to 18 months. This estimate is broken down by trade-secret category as follows:

• ML Model Architecture and Hyperparameter Configurations (6–8 months saved). The LumosEdge ensemble ML model architecture is the product of iterative research and experimentation spanning several years. Developing a comparable model architecture from scratch — including the neural-network topology design, ensemble integration framework, and the optimization of hyperparameter configurations — would require a skilled team of machine-learning engineers approximately 6 to 8 months of dedicated R&D effort. The hyperparameter configurations alone represent the distilled output of extensive computational experimentation, as identifying optimal hyperparameter settings for complex ensemble models is a time-intensive process involving substantial trial and error.

• Feature-Engineering Libraries (4–5 months saved). The 2,847 individually engineered features contained in Lumos's proprietary feature-engineering libraries represent an enormous body of domain-specific knowledge and data-science expertise. Each feature was developed to capture specific financial signals from alternative data sources and was subject to rigorous backtesting and validation. Independently replicating this library — including the identification, coding, testing, and validation of each feature — would require approximately 4 to 5 months even for a capable engineering team. Mr. Kapoor notes that the feature-engineering process is inherently iterative, as many candidate features are developed, tested, and ultimately discarded before a validated feature set is achieved.

• DataStream ETL Pipeline (3–4 months saved). Building a data ingestion pipeline with custom integrations for 43 separate alternative-data vendors is a substantial engineering undertaking. Each vendor integration requires understanding the vendor's data format, API specifications, authentication protocols, and data update cadences, as well as building robust error-handling, data-cleaning, and schema-validation logic. Mr. Kapoor estimates that developing a comparable pipeline from scratch would require 3 to 4 months, with a significant portion of that time consumed by the vendor-by-vendor integration and testing process.

• Product Roadmap intelligence. The value of the misappropriated product roadmap documents is more difficult to quantify in terms of development time saved, as the roadmap primarily provided strategic intelligence rather than executable technology. However, the roadmap gave NovaSpark significant insight into Lumos's planned product direction, allowing NovaSpark to prioritize features and capabilities that would directly compete with Lumos's upcoming releases. This strategic advantage, while not separately quantified in the time-saved analysis, supports the upper end of the unjust enrichment range and should be emphasized in settlement negotiations.

• Training Dataset (14.6 TB). The curated training dataset, by virtue of its size and the infrastructure required to transfer it, was largely not portable to NovaSpark in its entirety. However, Brindoval's forensic findings indicate that access to the DataStream pipeline's schema definitions and data-mapping configurations accelerated NovaSpark's own data acquisition strategy by providing a blueprint for the types of data to acquire and how to structure and label that data. This benefit is captured within the time-saved estimates for the ETL pipeline and feature-engineering categories above.

Valuation of Time Saved. Having established the range of development time saved at 14 to 18 months, the next step in the unjust enrichment analysis is to assign a dollar value to that time savings. We have adopted a cost-based approach, valuing the time saved by reference to the actual engineering and operational costs that NovaSpark would have incurred during the equivalent development period.

NovaSpark's $12 million seed funding round from Blackpine Capital Partners LP, led by managing partner Franklin Osei and closed on November 15, 2022, provides a market-based anchor for this analysis. Financial disclosures produced by NovaSpark in discovery indicate the following cost structure:

• NovaSpark's engineering team at the time of the seed round comprised the three founding engineers (Anand, Mallick, and Yun) plus four additional engineers hired between October and December 2022. Total engineering compensation (salaries, benefits, equity acceleration costs) averaged approximately $350,000 to $420,000 per month during the relevant period.

• Cloud computing infrastructure costs (Meridian Technologies and Aldermere Technologies) averaged approximately $65,000 to $85,000 per month.

• Data acquisition and alternative-data vendor licensing costs averaged approximately $45,000 to $75,000 per month.

• Other overhead (office lease, insurance, legal, administrative) averaged approximately $40,000 to $120,000 per month.

• Total monthly burn rate is therefore estimated at approximately $500,000 to $700,000 per month, with an average of approximately $620,000 per month at the low end and $690,000 per month at the high end when adjusted for the ramp-up in costs over the period.

Applying these burn rate estimates to the development-time range yields the following unjust enrichment calculation:

• Low-end estimate: 14 months × $620,000 average monthly burn = $8,680,000 (approximately $8.7 million).

• High-end estimate: 18 months × $690,000 adjusted average monthly burn = $12,420,000 (approximately $12.4 million).

These figures represent the costs that NovaSpark avoided incurring by using Lumos's trade secrets as a shortcut to product development. In other words, they measure the economic benefit — the unjust enrichment — conferred upon NovaSpark through the misappropriation.

We note that this analysis likely understates the full extent of NovaSpark's enrichment. In addition to the direct cost savings, NovaSpark's ability to reach market readiness sooner means earlier revenue generation, earlier achievement of product-market fit, and earlier investor traction. NovaSpark successfully closed its $12 million seed round on November 15, 2022 — only six weeks after Anand incorporated the company — a timeline that would be extraordinary for a genuinely de novo startup. The speed of this fundraising is circumstantial evidence that NovaSpark was able to demonstrate a level of technical maturity to investors that would not have been possible without the misappropriated Lumos technology. While we have not separately quantified this additional enrichment factor, it supports the upper end of the unjust enrichment range and should be highlighted in the mediation.

Unjust Enrichment Range: $8.7 million to $12.4 million.

C. Alternative Damages Theories

While the unjust enrichment theory is our primary damages framework, we briefly address two alternative theories for completeness, as well as the significant exemplary damages multiplier.

Actual Loss to Lumos. Proving Lumos's actual loss caused by the misappropriation is more complex and, at this stage, would likely require retention of an economic damages expert to model lost revenues and market-share erosion. Lumos's 2022 annual recurring revenue ("ARR") was approximately $67 million, and its customer base includes approximately 45 institutional clients — primarily hedge funds and proprietary trading desks. The competitive harm from NovaSpark's entry into the same market with a product built on Lumos's technology could, over time, affect Lumos's pricing power, customer retention, and new-customer acquisition. However, NovaSpark has not yet generated significant commercial revenues, and quantifying the future competitive impact involves a degree of speculation that makes this theory less reliable as a primary measure at this time. We recommend reserving the actual-loss theory for trial but not relying on it as the principal damages measure in the mediation. Should the case proceed to trial, we would recommend retaining a qualified economic damages expert to develop a lost-profits model.

Reasonable Royalty. As an alternative to actual loss and unjust enrichment, a reasonable royalty may be established by reference to a hypothetical licensing negotiation between Lumos and NovaSpark. This approach asks what the parties would have agreed upon as a license fee had they negotiated at arm's length for NovaSpark's use of the misappropriated technology. Given the comprehensive nature of the trade secrets at issue and their centrality to NovaSpark's entire product, a hypothetical license would need to account for the full value of the technology to NovaSpark's business. Based on comparable technology licensing transactions in the financial technology sector and applying the Georgia-Pacific-style factors adapted for trade-secret cases, we estimate that a reasonable royalty could fall in the range of $5 million to $10 million. However, this theory is inherently less precise than the unjust enrichment approach, and the hypothetical nature of the negotiation is a weakness — the parties would never have entered into a licensing arrangement under real-world conditions, as Lumos would not willingly license its core proprietary technology to a direct competitor founded by three of its former engineers.

Exemplary Damages. The exemplary damages provision of CUTSA, Cal. Civ. Code § 3426.3(c), permits an award of up to twice any compensatory damages if the court finds that the misappropriation was willful and malicious. The evidence in this case strongly supports such a finding:

• Anand commenced transferring Lumos materials to personal accounts four days before his resignation, while still an active employee subject to fiduciary duties.

• All three defendants coordinated their departures from Lumos and their immediate transition to NovaSpark, indicating a premeditated plan to misappropriate and exploit Lumos's technology.

• Anand incorporated NovaSpark just three days after his resignation, and Mallick commenced employment at NovaSpark just three days after her resignation — demonstrating that NovaSpark was formed specifically to exploit the misappropriated trade secrets.

• Yun admitted under oath that he used the Lumos notebooks in NovaSpark's product development.

• None of the defendants returned Lumos materials or confirmed deletion as required by their PIIAAs.

A finding of willful and malicious misappropriation would permit the court to double any compensatory damages award. Thus, at the low end of the unjust enrichment range ($8.7 million), exemplary damages could reach $17.4 million; at the high end ($12.4 million), exemplary damages could reach $24.8 million. While we do not expect a court to necessarily award the maximum exemplary damages, the threat of such an award is a powerful motivator for Defendants to settle. This exposure should be clearly communicated to Defendants and their counsel during the mediation.

D. Damages Summary Table

The following table summarizes the damages range under each theory:

These figures establish a substantial damages framework that provides Lumos with significant leverage in settlement negotiations. Even at the lowest end of the unjust enrichment range, without exemplary damages, the exposure far exceeds the settlement range we recommend below.

IV. Settlement Valuation and Risk Assessment

In recommending a settlement range, we have carefully considered both the factors that favor settlement and those that favor proceeding to trial. A responsible settlement valuation must account for the full spectrum of litigation risks, including the probability of success on the merits, the risk of an adverse ruling on key legal or factual issues, the practical challenges of collecting a judgment, and the time and cost required to reach a resolution through trial.

A. Litigation Risk Factors Favoring Settlement

The following factors weigh in favor of seeking a negotiated settlement rather than proceeding through trial:

1. Trial Uncertainty. While we believe the evidence is strong, litigation outcomes are inherently uncertain. A jury or judge could credit Defendants' assertion that certain elements of the technology they used were independently developed or constituted general industry knowledge that was already in the public domain. Defendants' counsel — Patricia Whitfield and Samuel Gerber of Whitfield, Raines & Gerber LLP — has been aggressive in asserting these defenses and has retained its own technical expert (not yet disclosed) to challenge the trade-secret status of some or all of the five categories. The independent-development defense is weakened substantially by Yun's deposition admission, but it has not been entirely eliminated with respect to all five categories of trade secrets. There is some risk that a fact-finder could find that certain categories — particularly the product roadmap, which by its nature has a limited shelf life — do not meet the full statutory definition of a trade secret or that the damages attributable to certain categories are less than we have estimated.

2. Collection Risk. NovaSpark's primary liquid asset is the proceeds of its $12 million seed round from Blackpine Capital Partners LP. Based on the burn rate analysis discussed above ($500,000–$700,000 per month), NovaSpark has likely consumed between $5.5 million and $7.7 million of these proceeds since the funding closed in November 2022. This means that as of August 2023, NovaSpark may have only $4.3 million to $6.5 million remaining, and this figure is declining each month. Every additional month of litigation increases the risk that NovaSpark's funds are exhausted before a judgment can be obtained and collected. The individual defendants — Anand, Mallick, and Yun — are software engineers and data scientists, not high-net-worth individuals with substantial personal assets. While we could potentially pursue piercing the corporate veil or personal liability theories, the practical collectability of a multi-million-dollar judgment from these individuals is limited. This collection risk is a significant factor favoring settlement at a discount from the full damages estimate.

3. Time and Cost of Continued Litigation. If the case proceeds through trial, we estimate an additional 12 to 18 months before a final judgment could be obtained, including the completion of fact discovery, expert discovery, summary judgment briefing, pretrial motions, and a trial of approximately 10 to 15 days. The estimated additional Calderwood & Harkness legal fees to take this case through trial are $800,000 to $1,200,000, on top of the approximately $1.0 million already billed to date. Additional expert costs — including a damages expert, supplemental forensic work by Brindoval, and potential rebuttal experts — would add to this total. These are significant expenditures that must be weighed against the incremental recovery that might be obtained at trial beyond a mediated settlement.

4. Injunctive Relief Is Paramount. From Lumos's perspective, the most important objective of this litigation is not monetary damages but rather stopping NovaSpark from continuing to use Lumos's proprietary technology and ensuring complete remediation. A negotiated settlement that includes a stipulated permanent injunction, verified forensic deletion, and ongoing audit rights achieves this objective with far greater certainty and speed than a contested trial. Even if Lumos were to obtain a permanent injunction after trial, enforcement and compliance monitoring would likely involve additional litigation. A consensual resolution with detailed remediation provisions reduces this friction.

5. Investor and Board Considerations. Lumos's investors have communicated their interest in an efficient resolution. Helen Trask, General Partner at Ridgeline Ventures (which led Lumos's Series B round), and Jonathan Holt, Managing Director at Northstar Growth Equity (which led Lumos's Series D round), both serve as board observers and have indicated a preference for resolving the matter without the distraction and negative publicity of a prolonged trial. CEO Priya Chandrasekaran, General Counsel Thomas Brightwell, and VP of Engineering Rachel Fong have each devoted substantial time to this litigation — time that could be better allocated to Lumos's product development and commercial operations. Prolonged litigation also creates a negative public narrative that could affect Lumos's relationships with current and prospective institutional clients, many of whom are sophisticated financial services firms with their own sensitivity to litigation risk.

B. Litigation Risk Factors Favoring Lumos at Trial

Notwithstanding the factors favoring settlement, Lumos's litigation position is strong, and this strength is an essential component of our settlement leverage. The following factors favor Lumos in the event the case proceeds to trial:

1. Strong Forensic Evidence. The Brindoval Preliminary Forensic Report provides clear, well-documented, and technically rigorous evidence that Lumos's proprietary materials were present on NovaSpark's corporate systems and each defendant's personal devices and cloud accounts. The forensic evidence is not circumstantial — it directly establishes the transfer, possession, and use of Lumos trade secrets. This type of forensic evidence is highly persuasive to both judges and juries, and it will be difficult for Defendants to rebut.

2. Yun's Deposition Admission. Derek Yun's admission that he used Lumos's Jupyter notebooks in NovaSpark's product development is a powerful piece of evidence that essentially concedes the element of use. This admission eliminates the independent-development defense with respect to at least the feature-engineering code and training scripts contained in the notebooks. In our experience, such a direct admission of use is relatively uncommon in trade-secret cases and gives Lumos a significant evidentiary advantage.

3. Coordinated Misconduct. The compressed timeline of the defendants' departures — Anand's September 30 resignation, Mallick's October 14 resignation, and Yun's October 20 resignation — combined with the incorporation of NovaSpark on October 3, 2022 (three days after Anand's departure) and the commencement of employment at NovaSpark within days of each resignation, demonstrates a coordinated scheme to leave Lumos and exploit its technology. This pattern of premeditated conduct is compelling evidence of wrongful intent and supports both liability and exemplary damages.

4. Breach of PIIAA. The individual defendants' obligations under their respective PIIAAs are clear and enforceable. The agreements contain unambiguous provisions regarding confidentiality, non-disclosure, assignment of inventions, and return of company materials. Each defendant is in plain breach of these obligations. The PIIAA claims provide an independent basis for liability that does not depend on trade-secret status and that is less susceptible to the defenses raised against the CUTSA claims.

5. Preliminary Injunction Already Granted. The Honorable David Rutherford granted Lumos's motion for a preliminary injunction on February 17, 2023, finding that Lumos had demonstrated a probability of prevailing on the merits of its trade-secret misappropriation claims. While a preliminary injunction ruling is not dispositive of the ultimate merits, it signals the Court's assessment that Lumos's claims are meritorious and that irreparable harm exists. This is a strong indicator of the likely outcome at trial and should be leveraged in the mediation.

6. Willfulness Supports Exemplary Damages. As discussed in Section III.C above, the evidence of willful and malicious misappropriation is strong. The prospect of exemplary damages up to twice the compensatory award represents a significant additional exposure for Defendants that should motivate settlement.

C. Recommended Settlement Range

Based on the foregoing analysis of damages estimates, litigation risks, collection concerns, and cost considerations, we recommend a target monetary settlement range of $4 million to $6 million, to be paid by NovaSpark and, if necessary, guaranteed by the individual defendants.

This range represents approximately 35% to 50% of the low-end unjust enrichment estimate of $8.7 million, and approximately 32% to 48% of the high-end estimate of $12.4 million. The rationale for this discount from the full unjust enrichment range is as follows:

• Collection Risk. Given NovaSpark's declining cash position and the limited personal assets of the individual defendants, there is a meaningful risk that a full judgment at the higher end of the damages range would not be collectible. A settlement in the $4–$6 million range is within the range of NovaSpark's likely remaining liquid assets and is therefore more likely to be actually collectible.

• Litigation Cost Avoidance. A settlement would avoid an estimated $800,000 to $1,200,000 in additional Calderwood & Harkness legal fees, plus additional expert costs. These savings enhance the net recovery to Lumos from a settlement in the recommended range.

• Certainty and Speed of Resolution. A settlement provides certainty of outcome and avoids the inherent risks of trial, including the possibility of an adverse ruling, a reduced damages award, or post-trial motions and appeals that could delay collection for years.

• Adequate Deterrence. A monetary payment in the $4–$6 million range, combined with the comprehensive non-monetary relief described below, provides meaningful deterrence against future misappropriation while recognizing the practical limitations on recovery.

The approximate midpoint of the recommended range is $4.35 million, which represents 50% of the low-end unjust enrichment estimate of $8.7 million and approximately 35.1% of the high-end estimate of $12.4 million. We believe this midpoint is a realistic and defensible settlement target given the totality of the circumstances.

D. Non-Monetary Settlement Objectives (Essential Terms)

In addition to the monetary component, the following non-monetary terms are essential to any settlement and should be treated as non-negotiable during the mediation:

1. Stipulated Permanent Injunction. NovaSpark, Anand, Mallick, and Yun shall be permanently enjoined from possessing, using, disclosing, or otherwise exploiting any Lumos trade secrets across all five categories identified in this memorandum. The stipulated injunction should be submitted to the Court as a consent order enforceable through contempt proceedings, providing Lumos with a direct enforcement mechanism in the event of any future violation.

2. Forensic Remediation. A complete, verified deletion and destruction of all Lumos proprietary materials from NovaSpark's corporate servers, all computing devices owned or controlled by the individual defendants, and all cloud storage accounts used by the individual defendants or NovaSpark. The remediation process shall be supervised by Brindoval Digital Forensics LLC, with technical oversight provided by Naveen Kapoor of Meridian Cybersecurity Advisors Inc. Brindoval shall issue a certification of completed remediation upon conclusion of the process.

3. Personal Covenants. Each individual defendant — Anand, Mallick, and Yun — shall personally covenant not to use, disclose, or otherwise exploit any Lumos trade secrets in any future employment, consulting arrangement, or business venture for a period of at least five years from the date of settlement. This provision extends beyond the PIIAA's three-year confidentiality obligation and reflects the severity of the misappropriation.

4. Audit Rights. Brindoval Digital Forensics LLC shall retain the right to conduct periodic audits of NovaSpark's computing systems and development environment for a period of at least 18 months following the settlement date to verify ongoing compliance with the stipulated permanent injunction and the forensic remediation requirements. NovaSpark shall be permitted to continue its business operations only using independently created technology or technology that is publicly available. Any audited materials found to contain or derive from Lumos trade secrets shall be subject to immediate deletion and may trigger contempt proceedings under the stipulated injunction.

5. Forensic Examiner Cost Allocation. We recommend that the costs of the forensic remediation and ongoing audit be allocated 60% to Defendants and 40% to Lumos, consistent with the cost-sharing arrangement established in the Court's March 8, 2023 stipulation and order appointing Brindoval as the neutral forensic examiner. Brindoval's total fees to date are $218,600. Under the existing 60/40 allocation, $131,160 is borne by Defendants/NovaSpark and $87,440 is borne by Lumos. Future forensic fees associated with the remediation and audit process should follow the same allocation, with Defendants bearing the majority share.

6. Mutual Non-Disparagement. Both sides should agree to a mutual non-disparagement provision prohibiting public statements that are derogatory or critical of the other party in connection with the litigation or the underlying events. This protects both Lumos's and Defendants' reputations in the marketplace.

7. Confidentiality of Settlement Terms. The financial terms of the settlement and the terms of the personal covenants should be subject to a confidentiality provision. However, the stipulated permanent injunction should be a public court filing, as it provides ongoing protection for Lumos and serves as a deterrent visible to others in the industry. Lumos's ability to reference the existence of the injunction — without disclosing the monetary terms — is important for its credibility with institutional clients and potential employees.

V. Mediation Strategy

The mediation is scheduled for September 6, 2023 before the Honorable Eleanor Voss (Ret.) at Tidewater Mediation Group, located at 160 West Santa Clara Street, Suite 1050, San Jose, California 95113. Judge Voss is a well-respected mediator with extensive experience in complex commercial and intellectual property disputes, and she is known for her ability to evaluate cases critically and to communicate realistic assessments to the parties.

We recommend the following strategic approach to the mediation:

Mediation Brief. A comprehensive mediation brief should be prepared and submitted to Judge Voss no later than August 30, 2023 — one week in advance of the mediation session — in accordance with Tidewater's standard protocols. The brief should present the full damages analysis set forth in this memorandum, including the unjust enrichment range of $8.7 million to $12.4 million and the exemplary damages potential of $17.4 million to $24.8 million. The brief should be supported by excerpts from the Brindoval Preliminary Forensic Report, the deposition transcripts (particularly Yun's admission), and Mr. Kapoor's technical assessment. We recommend submitting the brief on a confidential basis to the mediator only, with a redacted version available for the joint session if appropriate.

Opening Session. In the opening joint session, we recommend presenting the unjust enrichment analysis and the exemplary damages multiplier in a clear, direct manner calculated to maximize Lumos's leverage. The presentation should emphasize the strength of the forensic evidence, the damaging nature of Yun's deposition admission, the coordinated nature of the misconduct, and the Court's preliminary injunction ruling as an indicator of likely success on the merits. The goal of the opening session is to establish in the minds of Defendants and their counsel that Lumos's case is strong, that the damages exposure is substantial, and that the litigation risk for Defendants is significant.

Opening Demand. We recommend that Lumos open with a demand of $7.5 million in monetary compensation, plus all non-monetary terms described in Section IV.D above. This opening demand is within the unjust enrichment range and provides sufficient room to negotiate downward toward the target settlement range of $4 million to $6 million without signaling an immediate willingness to accept a figure at the lower end. The gap between the opening demand and the target range also allows for a structured negotiation process that respects the mediator's role and gives Defendants the sense that meaningful concessions are being made.

Key Arguments in Private Caucus. During private caucus sessions with Judge Voss, we should emphasize the following points:

• Collection risk. This is perhaps the most powerful argument for why Defendants should settle now. NovaSpark's cash is running out. Every month that passes without a settlement reduces the amount of money available to satisfy a judgment. If NovaSpark exhausts its funds before trial, Lumos will be left with a judgment against a shell entity and three individual defendants with limited personal assets. Settling now, while NovaSpark still has sufficient funds to make a meaningful payment, is in everyone's interest.

• Yun's deposition admission. This is a "smoking gun" that effectively eliminates the independent-development defense. We should ensure that Judge Voss understands the significance of Yun's admission and communicates its implications to Defendants' counsel during private caucuses. Defense counsel will have difficulty arguing that NovaSpark developed its technology independently when one of the three co-founders has admitted under oath to using Lumos's notebooks in NovaSpark's product development.

• Exemplary damages exposure. The threat of willful-and-malicious findings and a doubling of compensatory damages should be emphasized as a significant litigation risk for Defendants. Even if Defendants believe they can reduce the compensatory damages figure at trial, the exemplary damages multiplier creates an outsized exposure.

• Preliminary injunction ruling. The Court's February 17, 2023 order granting the preliminary injunction signals the Court's view on the merits and should be cited as evidence that the case is not a close call on liability.

Non-Monetary Terms. We should communicate clearly and early in the mediation that the non-monetary terms — stipulated permanent injunction, forensic remediation, personal covenants, and audit rights — are non-negotiable for Lumos. These terms are essential to achieving Lumos's core business objective of protecting its proprietary technology. We anticipate that Defendants may be more willing to accept these non-monetary terms than to agree to a larger monetary payment, which makes them a potentially useful point of leverage in the negotiation.

Mediation Fee. Tidewater Mediation Group charges a fee of $18,500 for the full-day mediation session, which is to be split equally between the parties ($9,250 per side).

VI. Fees and Costs Summary to Date

The following table summarizes the fees and costs incurred in connection with this matter through the date of this memorandum, as well as the projected allocation:

For reference, the applicable Calderwood & Harkness billing rates for this matter are as follows:

• Marcus Delgado, Responsible Partner: $1,150 per hour

• Catherine Bloom, Billing Partner: $1,050 per hour

• Wendy Schultz, Lead Associate: $725 per hour

• Omar Siddiqui, Working Associate: $575 per hour

Estimated Additional C&H Fees Through Trial (If No Settlement). If the mediation does not result in a settlement and the case proceeds through trial, we estimate that Calderwood & Harkness will incur an additional $800,000 to $1,200,000 in legal fees, depending on the complexity and duration of expert discovery, dispositive motion practice, and trial preparation. This estimate does not include additional Brindoval forensic costs for supplemental analysis, the retention of a damages expert for trial testimony, or potential rebuttal expert costs. When these additional expert costs are factored in, the total additional cost of proceeding through trial could reach $1.2 million to $1.6 million.

These projected costs further support the recommendation to seek a negotiated settlement at mediation, as the cost savings from avoiding trial enhance the effective net recovery to Lumos from a settlement in the recommended $4 million to $6 million range.

VII. Recommendation

Based on the analysis set forth in this memorandum, we recommend the following course of action:

1. Proceed with Mediation. Lumos should proceed with the September 6, 2023 mediation before the Honorable Eleanor Voss (Ret.) at Tidewater Mediation Group, seeking a settlement in the monetary range of $4 million to $6 million, together with all non-monetary terms identified in Section IV.D of this memorandum — specifically: a stipulated permanent injunction, comprehensive forensic remediation supervised by Brindoval, five-year personal covenants by each individual defendant, 18-month audit rights, mutual non-disparagement, and confidentiality of financial settlement terms.

2. Prepare and Submit Mediation Brief. A mediation brief presenting the full damages analysis and evidentiary record should be prepared and submitted to Judge Voss no later than August 30, 2023 (one week in advance of the mediation). We will circulate a draft of the mediation brief for partner review by August 23, 2023.

3. If Mediation Is Unsuccessful, Proceed to Trial. In the event that the mediation does not result in a settlement on acceptable terms, we recommend proceeding to trial. The evidentiary record supports a substantially larger recovery than the recommended settlement range, and the exemplary damages exposure provides significant additional motivation for Defendants to settle. If the mediation produces no resolution, we should promptly move to complete expert discovery and prepare for trial.

4. Client and Board Briefing. In advance of the mediation, we recommend scheduling a meeting with Lumos CEO Priya Chandrasekaran and General Counsel Thomas Brightwell to discuss the settlement parameters and obtain authorization to negotiate within the recommended range. We further recommend briefing Lumos's board representatives — Helen Trask (Ridgeline Ventures) and Jonathan Holt (Northstar Growth Equity) — on the mediation strategy and settlement range, to the extent authorized by management and consistent with board governance protocols.

VIII. Conclusion

The evidence developed through discovery to date — including the Brindoval forensic findings, the deposition testimony of each individual defendant, and the technical assessment of Meridian's Naveen Kapoor — supports a substantial recovery for Lumos Analytics Inc. The unjust enrichment conferred upon NovaSpark through its use of Lumos's misappropriated trade secrets is estimated at $8.7 million to $12.4 million, and the potential for exemplary damages amplifies this exposure to $17.4 million to $24.8 million. This damages framework provides Lumos with strong negotiating leverage in the upcoming mediation.

The recommended settlement target of $4 million to $6 million, when combined with the robust non-monetary protections described herein — the stipulated permanent injunction, forensic remediation, personal covenants, and audit rights — represents a favorable outcome that appropriately balances Lumos's potential recovery against the litigation risks, collection uncertainty, cost savings, and speed of resolution. Most importantly, a negotiated settlement with comprehensive injunctive and remediation terms achieves Lumos's primary business objective: the complete and verified elimination of its proprietary technology from NovaSpark's systems and the individual defendants' possession, with ongoing safeguards to ensure compliance.

We are available to discuss this analysis at your convenience and will proceed with preparation of the mediation brief upon your approval. Please do not hesitate to reach out with any questions, comments, or requested revisions.

Respectfully submitted,

CALDERWOOD & HARKNESS LLP

By: ________

Wendy Schultz Lead Associate

By: ________

Omar Siddiqui Working Associate

Enclosures:

1. Brindoval Digital Forensics LLC — Preliminary Forensic Report Summary (May 1, 2023)

2. Meridian Cybersecurity Advisors Inc. — Technical Assessment of Development Time Savings (July 12, 2023)

3. Deposition Transcript Excerpts — Rajiv Anand (May 15, 2023), Serena Mallick (June 7, 2023), and Derek Yun (June 22, 2023)

PRIVILEGED AND CONFIDENTIAL — ATTORNEY WORK PRODUCT This memorandum is protected by the attorney-client privilege and the work-product doctrine. Unauthorized disclosure is strictly prohibited.

Damages Theory	Low Estimate	High Estimate
Unjust Enrichment (Primary Theory)	$8,700,000	$12,400,000
Reasonable Royalty (Alternative)	$5,000,000	$10,000,000
Exemplary Damages (2× Unjust Enrichment)	$17,400,000	$24,800,000
Vendor / Firm	Total Fees	Allocation
Calderwood & Harkness LLP (through July 2023, approximate)	~$1,000,000 (estimated; final total will be confirmed with final billing reconciliation and will likely exceed this figure)	Borne entirely by Lumos
Brindoval Digital Forensics LLC	$218,600	60% Defendants ($131,160) / 40% Lumos ($87,440) per Court order dated March 8, 2023
Meridian Cybersecurity Advisors Inc.	$94,300	Borne entirely by Lumos
Tidewater Mediation Group (estimated)	$18,500	Split equally ($9,250 per side)

PRIVILEGED AND CONFIDENTIAL — ATTORNEY WORK PRODUCT
This memorandum constitutes attorney work product prepared in anticipation of litigation and is protected by the attorney-client privilege and the work-product doctrine. It is intended solely for the use of the attorneys and client identified herein and should not be disclosed to any third party without the express consent of the undersigned.
CALDERWOOD & HARKNESS LLP INTERNAL MEMORANDUM
FROM: Wendy Schultz, Lead Associate, and Omar Siddiqui, Working Associate
TO: Marcus Delgado, Responsible Partner, and Catherine Bloom, Billing Partner
DATE: August 15, 2023
RE: Damages Analysis and Settlement Valuation — Lumos Analytics Inc. v. NovaSpark Technologies Inc. et al., Case No. 23-CV-004871 (Superior Court of California, County of Santa Clara)
MATTER NUMBER: 1008-00005
I. Executive Summary
The purpose of this memorandum is to present a comprehensive damages analysis for the trade-secret misappropriation claims asserted by our client, Lumos Analytics Inc. ("Lumos"), against NovaSpark Technologies Inc. ("NovaSpark"), Rajiv Anand, Serena Mallick, and Derek Yun (collectively, "Defendants"), and to recommend a settlement range and negotiation strategy in advance of the mediation session scheduled for September 6, 2023 before the Honorable Eleanor Voss (Ret.) at Tidewater Mediation Group.
Bottom Line. Based on our analysis and the technical input received from Lumos's retained consultants, the unjust enrichment conferred upon NovaSpark through the misappropriation of Lumos's proprietary technology is estimated at $8.7 million to $12.4 million, reflecting approximately 14 to 18 months of accelerated development time that NovaSpark obtained by using Lumos's trade secrets rather than developing competing technology independently. We recommend a target settlement range of $4 million to $6 million in monetary compensation, together with a stipulated permanent injunction and a comprehensive forensic remediation protocol to be supervised by Brindoval Digital Forensics LLC ("Brindoval"), with technical oversight from Naveen Kapoor of Meridian Cybersecurity Advisors Inc. ("Meridian"). These non-monetary terms are essential to achieving Lumos's primary business objective of protecting its proprietary technology from ongoing exploitation.
The damages analysis set forth in this memorandum incorporates technical input from Naveen Kapoor, Senior Consultant at Meridian, Lumos's retained cybersecurity and technical consultant, and builds upon the findings of the Brindoval Preliminary Forensic Report dated May 1, 2023. The analysis is further informed by deposition testimony obtained from each of the three individual defendants, financial disclosures produced by NovaSpark in discovery, and publicly available information regarding NovaSpark's funding and operations.
The claims at issue in this litigation are as follows:
1. Misappropriation of trade secrets under the California Uniform Trade Secrets Act ("CUTSA"), Cal. Civ. Code §§ 3426–3426.11;
2. Breach of contract — specifically, each individual defendant's Proprietary Information and Inventions Assignment Agreement ("PIIAA") executed upon commencement of employment with Lumos;
3. Breach of the duty of loyalty owed by each individual defendant to Lumos during the period of their employment;
4. Unfair business practices in violation of California Business and Professions Code § 17200 et seq.; and
5. Conversion of Lumos proprietary materials and intellectual property.
This memorandum proceeds in seven parts. Part II summarizes the key evidence developed through discovery to date. Part III presents the damages analysis, including the primary unjust enrichment theory and alternative damages theories. Part IV provides a settlement valuation that accounts for litigation risks, collection concerns, and cost-of-litigation considerations. Part V outlines our recommended mediation strategy. Part VI summarizes fees and costs incurred to date and projected through trial. Part VII sets forth our recommendation for next steps.
II. Summary of Key Evidence Developed to Date
The evidence gathered through forensic investigation, document discovery, and deposition testimony provides a strong foundation for Lumos's claims. Below, we summarize the principal evidentiary categories and their significance to the damages analysis and settlement posture.
A. Forensic Evidence — Brindoval Preliminary Forensic Report (May 1, 2023)
Brindoval Digital Forensics LLC, led by Dr. Irene Chow, Managing Director, and Kevin Barr, Senior Forensic Analyst, conducted a comprehensive forensic investigation pursuant to the Court's March 8, 2023 stipulation and order. On May 1, 2023, Brindoval issued its Preliminary Forensic Report, which confirmed the presence of Lumos proprietary materials on NovaSpark's corporate servers, each individual defendant's personal computing devices, and each individual defendant's personal cloud storage accounts. The report is detailed and well-documented, and we believe it will be highly persuasive to both the mediator and, if necessary, the trier of fact.
The specific findings by defendant are summarized as follows:
Rajiv Anand. Brindoval confirmed that on September 26, 2022 — four days before his resignation from Lumos — Anand transferred approximately 3.2 gigabytes of files from his Lumos-issued workstation to a personal Hollowell Health IT account. The transferred files included Lumos's proprietary model architecture specifications and hyperparameter configuration files for the LumosEdge ensemble machine-learning model, which constitutes the core of Lumos's commercial product offering. Additionally, on September 28, 2022, Anand forwarded a 37-page confidential document titled "LumosEdge 3.0 Product Roadmap" in PDF format to his personal email address at r.anand.personal@vaultmail.com. The Product Roadmap contained Lumos's strategic plans for the 2023–2025 product development cycle, including unreleased feature specifications and competitive positioning analysis. Anand did not have authorization to transfer any of these materials to personal accounts, and Lumos's information security policies expressly prohibit such transfers.
Serena Mallick. On October 10, 2022 — four days before her resignation from Lumos — Mallick cloned the entirety of the DataStream ETL pipeline repository, comprising approximately 780 megabytes of code, configuration files, and data-schema definitions, to a personal external USB drive bearing the serial number WD-4837291. The DataStream ETL pipeline is one of Lumos's most significant proprietary assets, as it manages the ingestion, transformation, and loading of data from 43 separate alternative-data vendor integrations. The pipeline took Lumos's engineering team approximately three years to develop and refine, and it represents a substantial competitive advantage in the alternative-data analytics market. Brindoval confirmed that the USB drive was subsequently connected to a computing device registered to NovaSpark's corporate network.
Derek Yun. On October 18, 2022 — two days before his resignation from Lumos — Yun downloaded 47 proprietary Jupyter notebooks from Lumos's internal development environment to his personal Larkspur Technologies account. These notebooks contained Lumos's proprietary feature-engineering code and model-training scripts, which together embody the accumulated work of Lumos's data science team over a period of approximately four years. The notebooks contain code for generating and validating 2,847 individually engineered features used in Lumos's financial signal-generation models. Brindoval's analysis of metadata and access logs confirmed that these notebooks were subsequently accessed and utilized within NovaSpark's development environment.
Across all three defendants, Brindoval's metadata and timestamp analysis corroborates the exfiltration timeline set forth above. The report demonstrates a coordinated pattern of data extraction occurring within a compressed timeframe — from September 26 through October 18, 2022 — immediately preceding each defendant's resignation from Lumos. The forensic evidence further confirms that the misappropriated materials were not merely downloaded and retained but were actively accessed and incorporated into NovaSpark's development workflow. This finding is critical because it establishes not only the act of misappropriation but also the actual use of Lumos's trade secrets in developing a competing product.
B. Deposition Testimony
We conducted depositions of each of the three individual defendants during May and June 2023. The testimony obtained is summarized below and includes several significant admissions that substantially strengthen Lumos's case.
Rajiv Anand (May 15, 2023). Anand acknowledged transferring files from his Lumos workstation to his personal Hollowell Health IT account on September 26, 2022, but characterized the transferred materials as "general reference materials" and "industry knowledge" that he believed he was entitled to retain. This characterization is directly contradicted by the forensic evidence, which confirms that the transferred files included Lumos's proprietary model architecture specifications and hyperparameter configurations — materials that are clearly identified as confidential in Lumos's information classification system and that bear proprietary markings. Anand further admitted that he coordinated his departure from Lumos with Mallick and Yun, and that the three of them had discussed forming a new venture together prior to resigning. Critically, corporate records produced in discovery confirm that Anand incorporated NovaSpark Technologies Inc. with the California Secretary of State on October 3, 2022 — just three days after his September 30, 2022 resignation from Lumos. This timeline strongly suggests premeditation and planning while still employed by Lumos.
Serena Mallick (June 7, 2023). Mallick admitted during her deposition that she copied the DataStream repository to the external USB drive (serial number WD-4837291) on October 10, 2022, but claimed that she did so merely to "retain personal notes" that she had created during her employment at Lumos. This explanation is not credible. The DataStream ETL pipeline repository consists of approximately 780 megabytes of complex production code, configuration files, and schema definitions covering data ingestion from 43 alternative-data vendor integrations. It is not a collection of personal notes. The repository represents a comprehensive, integrated codebase that would be of enormous value to any competitor seeking to rapidly establish alternative-data ingestion capabilities. Mallick further confirmed that she commenced employment at NovaSpark as Vice President of Engineering on October 17, 2022, just three days after her October 14, 2022 resignation from Lumos. She did not take any period of separation between the two positions.
Derek Yun (June 22, 2023). Yun's deposition yielded what we regard as the most significant evidentiary admission in the case. Yun acknowledged downloading 47 proprietary Jupyter notebooks from Lumos's internal systems to his personal Larkspur Technologies account on October 18, 2022. More importantly, and unlike his co-defendants, Yun went further and admitted that he subsequently used the downloaded notebooks in NovaSpark's product development. Specifically, Yun testified that the notebooks served as a "starting point" and "reference framework" for developing NovaSpark's competing feature-engineering capabilities. This admission is devastating to NovaSpark's independent-development defense and establishes actual use of Lumos trade secrets in the development of a competing product — a critical element under CUTSA. We expect Defendants' counsel will attempt to minimize the significance of this admission, but the deposition transcript is clear and unambiguous.
PIIAA Obligations. Each of the three individual defendants executed Lumos's standard Proprietary Information and Inventions Assignment Agreement upon commencement of their employment. The PIIAA contains the following material restrictive covenant provisions:
(a) Assignment of Inventions. All inventions, discoveries, developments, and improvements conceived or made by the employee during employment, or within twelve months thereafter if related to Lumos's business, are assigned to Lumos.
(b) Confidentiality and Non-Disclosure. The employee is obligated to maintain the confidentiality of all Lumos proprietary information and trade secrets, with this obligation surviving termination of employment for a period of three years.
(c) Non-Solicitation. The employee is prohibited from soliciting Lumos employees for a period of twelve months following separation from employment.
(d) Covenant Not to Use or Disclose Trade Secrets. The employee covenants not to use or disclose Lumos trade secrets at any time, without limitation, consistent with California law governing the protection of trade secrets.
None of the three individual defendants returned company materials or confirmed the deletion of Lumos proprietary information from their personal devices and accounts upon their departure from Lumos, as required by Section 7 of the PIIAA. This failure to comply with the return-of-materials obligation is itself a breach of contract and further evidences the intent to retain and use Lumos's proprietary materials at NovaSpark.
C. Trade Secrets at Issue — Five Categories
Through the course of this litigation, we have identified five categories of Lumos trade secrets that were misappropriated by Defendants. These categories have been specifically identified in our trade-secret disclosure statement filed with the Court on January 30, 2023, as supplemented on April 14, 2023, and they form the basis for our damages analysis:
1. LumosEdge Ensemble ML Model Architecture. This includes the proprietary neural-network topology, ensemble model configuration, and hyperparameter configurations that define the core algorithmic engine of Lumos's commercial product. The model architecture reflects years of iterative development, extensive experimentation, and significant investment in research and development. Lumos has invested approximately $14.2 million in R&D costs over the past four fiscal years to develop and refine this architecture.
2. Proprietary Feature-Engineering Libraries. These libraries contain 2,847 individually engineered features for financial signal generation, representing the accumulated expertise of Lumos's data science team. Each feature was developed, tested, and validated through a rigorous multi-stage process involving backtesting across historical market conditions. The feature libraries are among Lumos's most closely guarded assets, as they directly influence the predictive accuracy and commercial value of the LumosEdge product.
3. DataStream ETL Pipeline Codebase and Data-Schema Definitions. The DataStream ETL pipeline manages the ingestion, transformation, and loading of data from 43 alternative-data vendors. The pipeline is a complex, integrated system that took Lumos's engineering team approximately three years to develop. It includes custom connectors for each vendor's API, data-cleaning and normalization routines, schema validation logic, and error-handling protocols. The pipeline's vendor integrations are a significant barrier to entry in the alternative-data analytics market.
4. Curated and Labeled Training Dataset. Lumos has assembled and curated a training dataset of approximately 14.6 terabytes of historical financial and alternative data. This dataset has been labeled and structured for use in training and validating Lumos's machine-learning models. While the underlying raw data may be obtained from third-party sources, the curation, labeling, structuring, and quality-assurance processes applied to the dataset represent substantial independent value and constitute trade-secret-protectable work product.
5. Confidential Product Roadmap Documents (2023–2025). These documents include the 37-page "LumosEdge 3.0 Product Roadmap" PDF that Anand forwarded to his personal email, as well as related strategic planning documents. The roadmap contains unreleased product specifications, feature prioritization analysis, competitive intelligence, market positioning strategy, and planned vendor partnerships. Disclosure of this information to a competitor provides a significant strategic advantage by revealing Lumos's planned direction of development.
Each of these categories satisfies the statutory definition of a trade secret under CUTSA, Cal. Civ. Code § 3426.1(d), as information that derives independent economic value from not being generally known or readily ascertainable and that has been the subject of reasonable efforts to maintain its secrecy. Lumos maintains comprehensive information security measures, including role-based access controls, encryption, secure development environments, and the PIIAA obligations discussed above.
III. Damages Analysis
A. Applicable Legal Framework
The California Uniform Trade Secrets Act provides for multiple measures of damages for trade-secret misappropriation. Under Cal. Civ. Code § 3426.3, a plaintiff may recover damages measured by one or more of the following:
(i) Actual Loss. The actual loss to the plaintiff caused by the misappropriation, including both direct losses and consequential damages. Cal. Civ. Code § 3426.3(a).
(ii) Unjust Enrichment. The unjust enrichment caused by the misappropriation that is not taken into account in computing actual loss. This measure allows recovery of the benefit conferred upon the defendant through the wrongful use of the plaintiff's trade secrets, even where the plaintiff's own losses are difficult to quantify. Cal. Civ. Code § 3426.3(a).
(iii) Reasonable Royalty. In lieu of damages measured by any other method, the court may order payment of a reasonable royalty for no longer than the period of time the use could have been prohibited. Cal. Civ. Code § 3426.3(b). This measure is typically applied as a floor or alternative when actual loss and unjust enrichment are difficult to establish.
Additionally, Cal. Civ. Code § 3426.3(c) provides that if willful and malicious misappropriation is proven, the court may award exemplary damages in an amount not exceeding twice any compensatory damages award. This provision serves both a punitive and a deterrent function and is an important component of Lumos's overall damages exposure framework for purposes of settlement negotiations.
For purposes of this analysis, we focus primarily on the unjust enrichment theory as the most readily quantifiable and persuasive measure of damages given the evidence currently available. The unjust enrichment approach is well-suited to this case because: (a) the evidence clearly establishes that NovaSpark used Lumos's trade secrets to accelerate its product development, (b) the development-time savings can be quantified with reasonable certainty based on technical expert analysis, and (c) NovaSpark's financial records provide a market-based anchor for valuing the time saved. We discuss alternative theories — actual loss and reasonable royalty — below, as well as the exemplary damages multiplier.
B. Unjust Enrichment Analysis — Accelerated Time-to-Market
Methodology. The core of our damages analysis rests on quantifying the development time that NovaSpark saved by misappropriating and using Lumos's proprietary materials, rather than developing competing technology independently through its own research and engineering efforts. This "accelerated time-to-market" approach is well-established in trade-secret litigation and has been accepted by California courts as a basis for calculating unjust enrichment. See, e.g., Sargent Fletcher, Inc. v. Able Corp., 110 Cal. App. 4th 1658, 1670 (2003) (unjust enrichment may be measured by the "development costs the defendant avoided through misappropriation"); Ajaxo Inc. v. ETrade Group, Inc.*, 135 Cal. App. 4th 21, 57–58 (2005).
Our retained technical consultant, Naveen Kapoor, Senior Consultant at Meridian Cybersecurity Advisors Inc., conducted an independent assessment of the development timeline that NovaSpark would have required to replicate the functionality provided by the five categories of misappropriated Lumos trade secrets, assuming NovaSpark had proceeded entirely through independent development without access to any of Lumos's proprietary materials. Mr. Kapoor's assessment is based on his review of the Brindoval forensic findings, his independent analysis of the complexity of the misappropriated technologies, his experience in evaluating the development timelines for comparable machine-learning and data-engineering systems in the financial technology sector, and publicly available information regarding NovaSpark's team composition and technical capabilities.
Time Saved — Estimated at 14 to 18 Months. Based on Mr. Kapoor's technical assessment, the total development time saved by NovaSpark through its use of Lumos's proprietary materials is estimated at 14 to 18 months. This estimate is broken down by trade-secret category as follows:
• ML Model Architecture and Hyperparameter Configurations (6–8 months saved). The LumosEdge ensemble ML model architecture is the product of iterative research and experimentation spanning several years. Developing a comparable model architecture from scratch — including the neural-network topology design, ensemble integration framework, and the optimization of hyperparameter configurations — would require a skilled team of machine-learning engineers approximately 6 to 8 months of dedicated R&D effort. The hyperparameter configurations alone represent the distilled output of extensive computational experimentation, as identifying optimal hyperparameter settings for complex ensemble models is a time-intensive process involving substantial trial and error.
• Feature-Engineering Libraries (4–5 months saved). The 2,847 individually engineered features contained in Lumos's proprietary feature-engineering libraries represent an enormous body of domain-specific knowledge and data-science expertise. Each feature was developed to capture specific financial signals from alternative data sources and was subject to rigorous backtesting and validation. Independently replicating this library — including the identification, coding, testing, and validation of each feature — would require approximately 4 to 5 months even for a capable engineering team. Mr. Kapoor notes that the feature-engineering process is inherently iterative, as many candidate features are developed, tested, and ultimately discarded before a validated feature set is achieved.
• DataStream ETL Pipeline (3–4 months saved). Building a data ingestion pipeline with custom integrations for 43 separate alternative-data vendors is a substantial engineering undertaking. Each vendor integration requires understanding the vendor's data format, API specifications, authentication protocols, and data update cadences, as well as building robust error-handling, data-cleaning, and schema-validation logic. Mr. Kapoor estimates that developing a comparable pipeline from scratch would require 3 to 4 months, with a significant portion of that time consumed by the vendor-by-vendor integration and testing process.
• Product Roadmap intelligence. The value of the misappropriated product roadmap documents is more difficult to quantify in terms of development time saved, as the roadmap primarily provided strategic intelligence rather than executable technology. However, the roadmap gave NovaSpark significant insight into Lumos's planned product direction, allowing NovaSpark to prioritize features and capabilities that would directly compete with Lumos's upcoming releases. This strategic advantage, while not separately quantified in the time-saved analysis, supports the upper end of the unjust enrichment range and should be emphasized in settlement negotiations.
• Training Dataset (14.6 TB). The curated training dataset, by virtue of its size and the infrastructure required to transfer it, was largely not portable to NovaSpark in its entirety. However, Brindoval's forensic findings indicate that access to the DataStream pipeline's schema definitions and data-mapping configurations accelerated NovaSpark's own data acquisition strategy by providing a blueprint for the types of data to acquire and how to structure and label that data. This benefit is captured within the time-saved estimates for the ETL pipeline and feature-engineering categories above.
Valuation of Time Saved. Having established the range of development time saved at 14 to 18 months, the next step in the unjust enrichment analysis is to assign a dollar value to that time savings. We have adopted a cost-based approach, valuing the time saved by reference to the actual engineering and operational costs that NovaSpark would have incurred during the equivalent development period.
NovaSpark's $12 million seed funding round from Blackpine Capital Partners LP, led by managing partner Franklin Osei and closed on November 15, 2022, provides a market-based anchor for this analysis. Financial disclosures produced by NovaSpark in discovery indicate the following cost structure:
• NovaSpark's engineering team at the time of the seed round comprised the three founding engineers (Anand, Mallick, and Yun) plus four additional engineers hired between October and December 2022. Total engineering compensation (salaries, benefits, equity acceleration costs) averaged approximately $350,000 to $420,000 per month during the relevant period.
• Cloud computing infrastructure costs (Meridian Technologies and Aldermere Technologies) averaged approximately $65,000 to $85,000 per month.
• Data acquisition and alternative-data vendor licensing costs averaged approximately $45,000 to $75,000 per month.
• Other overhead (office lease, insurance, legal, administrative) averaged approximately $40,000 to $120,000 per month.
• Total monthly burn rate is therefore estimated at approximately $500,000 to $700,000 per month, with an average of approximately $620,000 per month at the low end and $690,000 per month at the high end when adjusted for the ramp-up in costs over the period.
Applying these burn rate estimates to the development-time range yields the following unjust enrichment calculation:
• Low-end estimate: 14 months × $620,000 average monthly burn = $8,680,000 (approximately $8.7 million).
• High-end estimate: 18 months × $690,000 adjusted average monthly burn = $12,420,000 (approximately $12.4 million).
These figures represent the costs that NovaSpark avoided incurring by using Lumos's trade secrets as a shortcut to product development. In other words, they measure the economic benefit — the unjust enrichment — conferred upon NovaSpark through the misappropriation.
We note that this analysis likely understates the full extent of NovaSpark's enrichment. In addition to the direct cost savings, NovaSpark's ability to reach market readiness sooner means earlier revenue generation, earlier achievement of product-market fit, and earlier investor traction. NovaSpark successfully closed its $12 million seed round on November 15, 2022 — only six weeks after Anand incorporated the company — a timeline that would be extraordinary for a genuinely de novo startup. The speed of this fundraising is circumstantial evidence that NovaSpark was able to demonstrate a level of technical maturity to investors that would not have been possible without the misappropriated Lumos technology. While we have not separately quantified this additional enrichment factor, it supports the upper end of the unjust enrichment range and should be highlighted in the mediation.
Unjust Enrichment Range: $8.7 million to $12.4 million.
C. Alternative Damages Theories
While the unjust enrichment theory is our primary damages framework, we briefly address two alternative theories for completeness, as well as the significant exemplary damages multiplier.
Actual Loss to Lumos. Proving Lumos's actual loss caused by the misappropriation is more complex and, at this stage, would likely require retention of an economic damages expert to model lost revenues and market-share erosion. Lumos's 2022 annual recurring revenue ("ARR") was approximately $67 million, and its customer base includes approximately 45 institutional clients — primarily hedge funds and proprietary trading desks. The competitive harm from NovaSpark's entry into the same market with a product built on Lumos's technology could, over time, affect Lumos's pricing power, customer retention, and new-customer acquisition. However, NovaSpark has not yet generated significant commercial revenues, and quantifying the future competitive impact involves a degree of speculation that makes this theory less reliable as a primary measure at this time. We recommend reserving the actual-loss theory for trial but not relying on it as the principal damages measure in the mediation. Should the case proceed to trial, we would recommend retaining a qualified economic damages expert to develop a lost-profits model.
Reasonable Royalty. As an alternative to actual loss and unjust enrichment, a reasonable royalty may be established by reference to a hypothetical licensing negotiation between Lumos and NovaSpark. This approach asks what the parties would have agreed upon as a license fee had they negotiated at arm's length for NovaSpark's use of the misappropriated technology. Given the comprehensive nature of the trade secrets at issue and their centrality to NovaSpark's entire product, a hypothetical license would need to account for the full value of the technology to NovaSpark's business. Based on comparable technology licensing transactions in the financial technology sector and applying the Georgia-Pacific-style factors adapted for trade-secret cases, we estimate that a reasonable royalty could fall in the range of $5 million to $10 million. However, this theory is inherently less precise than the unjust enrichment approach, and the hypothetical nature of the negotiation is a weakness — the parties would never have entered into a licensing arrangement under real-world conditions, as Lumos would not willingly license its core proprietary technology to a direct competitor founded by three of its former engineers.
Exemplary Damages. The exemplary damages provision of CUTSA, Cal. Civ. Code § 3426.3(c), permits an award of up to twice any compensatory damages if the court finds that the misappropriation was willful and malicious. The evidence in this case strongly supports such a finding:
• Anand commenced transferring Lumos materials to personal accounts four days before his resignation, while still an active employee subject to fiduciary duties.
• All three defendants coordinated their departures from Lumos and their immediate transition to NovaSpark, indicating a premeditated plan to misappropriate and exploit Lumos's technology.
• Anand incorporated NovaSpark just three days after his resignation, and Mallick commenced employment at NovaSpark just three days after her resignation — demonstrating that NovaSpark was formed specifically to exploit the misappropriated trade secrets.
• Yun admitted under oath that he used the Lumos notebooks in NovaSpark's product development.
• None of the defendants returned Lumos materials or confirmed deletion as required by their PIIAAs.
A finding of willful and malicious misappropriation would permit the court to double any compensatory damages award. Thus, at the low end of the unjust enrichment range ($8.7 million), exemplary damages could reach $17.4 million; at the high end ($12.4 million), exemplary damages could reach $24.8 million. While we do not expect a court to necessarily award the maximum exemplary damages, the threat of such an award is a powerful motivator for Defendants to settle. This exposure should be clearly communicated to Defendants and their counsel during the mediation.
D. Damages Summary Table
The following table summarizes the damages range under each theory:
These figures establish a substantial damages framework that provides Lumos with significant leverage in settlement negotiations. Even at the lowest end of the unjust enrichment range, without exemplary damages, the exposure far exceeds the settlement range we recommend below.
IV. Settlement Valuation and Risk Assessment
In recommending a settlement range, we have carefully considered both the factors that favor settlement and those that favor proceeding to trial. A responsible settlement valuation must account for the full spectrum of litigation risks, including the probability of success on the merits, the risk of an adverse ruling on key legal or factual issues, the practical challenges of collecting a judgment, and the time and cost required to reach a resolution through trial.
A. Litigation Risk Factors Favoring Settlement
The following factors weigh in favor of seeking a negotiated settlement rather than proceeding through trial:
1. Trial Uncertainty. While we believe the evidence is strong, litigation outcomes are inherently uncertain. A jury or judge could credit Defendants' assertion that certain elements of the technology they used were independently developed or constituted general industry knowledge that was already in the public domain. Defendants' counsel — Patricia Whitfield and Samuel Gerber of Whitfield, Raines & Gerber LLP — has been aggressive in asserting these defenses and has retained its own technical expert (not yet disclosed) to challenge the trade-secret status of some or all of the five categories. The independent-development defense is weakened substantially by Yun's deposition admission, but it has not been entirely eliminated with respect to all five categories of trade secrets. There is some risk that a fact-finder could find that certain categories — particularly the product roadmap, which by its nature has a limited shelf life — do not meet the full statutory definition of a trade secret or that the damages attributable to certain categories are less than we have estimated.
2. Collection Risk. NovaSpark's primary liquid asset is the proceeds of its $12 million seed round from Blackpine Capital Partners LP. Based on the burn rate analysis discussed above ($500,000–$700,000 per month), NovaSpark has likely consumed between $5.5 million and $7.7 million of these proceeds since the funding closed in November 2022. This means that as of August 2023, NovaSpark may have only $4.3 million to $6.5 million remaining, and this figure is declining each month. Every additional month of litigation increases the risk that NovaSpark's funds are exhausted before a judgment can be obtained and collected. The individual defendants — Anand, Mallick, and Yun — are software engineers and data scientists, not high-net-worth individuals with substantial personal assets. While we could potentially pursue piercing the corporate veil or personal liability theories, the practical collectability of a multi-million-dollar judgment from these individuals is limited. This collection risk is a significant factor favoring settlement at a discount from the full damages estimate.
3. Time and Cost of Continued Litigation. If the case proceeds through trial, we estimate an additional 12 to 18 months before a final judgment could be obtained, including the completion of fact discovery, expert discovery, summary judgment briefing, pretrial motions, and a trial of approximately 10 to 15 days. The estimated additional Calderwood & Harkness legal fees to take this case through trial are $800,000 to $1,200,000, on top of the approximately $1.0 million already billed to date. Additional expert costs — including a damages expert, supplemental forensic work by Brindoval, and potential rebuttal experts — would add to this total. These are significant expenditures that must be weighed against the incremental recovery that might be obtained at trial beyond a mediated settlement.
4. Injunctive Relief Is Paramount. From Lumos's perspective, the most important objective of this litigation is not monetary damages but rather stopping NovaSpark from continuing to use Lumos's proprietary technology and ensuring complete remediation. A negotiated settlement that includes a stipulated permanent injunction, verified forensic deletion, and ongoing audit rights achieves this objective with far greater certainty and speed than a contested trial. Even if Lumos were to obtain a permanent injunction after trial, enforcement and compliance monitoring would likely involve additional litigation. A consensual resolution with detailed remediation provisions reduces this friction.
5. Investor and Board Considerations. Lumos's investors have communicated their interest in an efficient resolution. Helen Trask, General Partner at Ridgeline Ventures (which led Lumos's Series B round), and Jonathan Holt, Managing Director at Northstar Growth Equity (which led Lumos's Series D round), both serve as board observers and have indicated a preference for resolving the matter without the distraction and negative publicity of a prolonged trial. CEO Priya Chandrasekaran, General Counsel Thomas Brightwell, and VP of Engineering Rachel Fong have each devoted substantial time to this litigation — time that could be better allocated to Lumos's product development and commercial operations. Prolonged litigation also creates a negative public narrative that could affect Lumos's relationships with current and prospective institutional clients, many of whom are sophisticated financial services firms with their own sensitivity to litigation risk.
B. Litigation Risk Factors Favoring Lumos at Trial
Notwithstanding the factors favoring settlement, Lumos's litigation position is strong, and this strength is an essential component of our settlement leverage. The following factors favor Lumos in the event the case proceeds to trial:
1. Strong Forensic Evidence. The Brindoval Preliminary Forensic Report provides clear, well-documented, and technically rigorous evidence that Lumos's proprietary materials were present on NovaSpark's corporate systems and each defendant's personal devices and cloud accounts. The forensic evidence is not circumstantial — it directly establishes the transfer, possession, and use of Lumos trade secrets. This type of forensic evidence is highly persuasive to both judges and juries, and it will be difficult for Defendants to rebut.
2. Yun's Deposition Admission. Derek Yun's admission that he used Lumos's Jupyter notebooks in NovaSpark's product development is a powerful piece of evidence that essentially concedes the element of use. This admission eliminates the independent-development defense with respect to at least the feature-engineering code and training scripts contained in the notebooks. In our experience, such a direct admission of use is relatively uncommon in trade-secret cases and gives Lumos a significant evidentiary advantage.
3. Coordinated Misconduct. The compressed timeline of the defendants' departures — Anand's September 30 resignation, Mallick's October 14 resignation, and Yun's October 20 resignation — combined with the incorporation of NovaSpark on October 3, 2022 (three days after Anand's departure) and the commencement of employment at NovaSpark within days of each resignation, demonstrates a coordinated scheme to leave Lumos and exploit its technology. This pattern of premeditated conduct is compelling evidence of wrongful intent and supports both liability and exemplary damages.
4. Breach of PIIAA. The individual defendants' obligations under their respective PIIAAs are clear and enforceable. The agreements contain unambiguous provisions regarding confidentiality, non-disclosure, assignment of inventions, and return of company materials. Each defendant is in plain breach of these obligations. The PIIAA claims provide an independent basis for liability that does not depend on trade-secret status and that is less susceptible to the defenses raised against the CUTSA claims.
5. Preliminary Injunction Already Granted. The Honorable David Rutherford granted Lumos's motion for a preliminary injunction on February 17, 2023, finding that Lumos had demonstrated a probability of prevailing on the merits of its trade-secret misappropriation claims. While a preliminary injunction ruling is not dispositive of the ultimate merits, it signals the Court's assessment that Lumos's claims are meritorious and that irreparable harm exists. This is a strong indicator of the likely outcome at trial and should be leveraged in the mediation.
6. Willfulness Supports Exemplary Damages. As discussed in Section III.C above, the evidence of willful and malicious misappropriation is strong. The prospect of exemplary damages up to twice the compensatory award represents a significant additional exposure for Defendants that should motivate settlement.
C. Recommended Settlement Range
Based on the foregoing analysis of damages estimates, litigation risks, collection concerns, and cost considerations, we recommend a target monetary settlement range of $4 million to $6 million, to be paid by NovaSpark and, if necessary, guaranteed by the individual defendants.
This range represents approximately 35% to 50% of the low-end unjust enrichment estimate of $8.7 million, and approximately 32% to 48% of the high-end estimate of $12.4 million. The rationale for this discount from the full unjust enrichment range is as follows:
• Collection Risk. Given NovaSpark's declining cash position and the limited personal assets of the individual defendants, there is a meaningful risk that a full judgment at the higher end of the damages range would not be collectible. A settlement in the $4–$6 million range is within the range of NovaSpark's likely remaining liquid assets and is therefore more likely to be actually collectible.
• Litigation Cost Avoidance. A settlement would avoid an estimated $800,000 to $1,200,000 in additional Calderwood & Harkness legal fees, plus additional expert costs. These savings enhance the net recovery to Lumos from a settlement in the recommended range.
• Certainty and Speed of Resolution. A settlement provides certainty of outcome and avoids the inherent risks of trial, including the possibility of an adverse ruling, a reduced damages award, or post-trial motions and appeals that could delay collection for years.
• Adequate Deterrence. A monetary payment in the $4–$6 million range, combined with the comprehensive non-monetary relief described below, provides meaningful deterrence against future misappropriation while recognizing the practical limitations on recovery.
The approximate midpoint of the recommended range is $4.35 million, which represents 50% of the low-end unjust enrichment estimate of $8.7 million and approximately 35.1% of the high-end estimate of $12.4 million. We believe this midpoint is a realistic and defensible settlement target given the totality of the circumstances.
D. Non-Monetary Settlement Objectives (Essential Terms)
In addition to the monetary component, the following non-monetary terms are essential to any settlement and should be treated as non-negotiable during the mediation:
1. Stipulated Permanent Injunction. NovaSpark, Anand, Mallick, and Yun shall be permanently enjoined from possessing, using, disclosing, or otherwise exploiting any Lumos trade secrets across all five categories identified in this memorandum. The stipulated injunction should be submitted to the Court as a consent order enforceable through contempt proceedings, providing Lumos with a direct enforcement mechanism in the event of any future violation.
2. Forensic Remediation. A complete, verified deletion and destruction of all Lumos proprietary materials from NovaSpark's corporate servers, all computing devices owned or controlled by the individual defendants, and all cloud storage accounts used by the individual defendants or NovaSpark. The remediation process shall be supervised by Brindoval Digital Forensics LLC, with technical oversight provided by Naveen Kapoor of Meridian Cybersecurity Advisors Inc. Brindoval shall issue a certification of completed remediation upon conclusion of the process.
3. Personal Covenants. Each individual defendant — Anand, Mallick, and Yun — shall personally covenant not to use, disclose, or otherwise exploit any Lumos trade secrets in any future employment, consulting arrangement, or business venture for a period of at least five years from the date of settlement. This provision extends beyond the PIIAA's three-year confidentiality obligation and reflects the severity of the misappropriation.
4. Audit Rights. Brindoval Digital Forensics LLC shall retain the right to conduct periodic audits of NovaSpark's computing systems and development environment for a period of at least 18 months following the settlement date to verify ongoing compliance with the stipulated permanent injunction and the forensic remediation requirements. NovaSpark shall be permitted to continue its business operations only using independently created technology or technology that is publicly available. Any audited materials found to contain or derive from Lumos trade secrets shall be subject to immediate deletion and may trigger contempt proceedings under the stipulated injunction.
5. Forensic Examiner Cost Allocation. We recommend that the costs of the forensic remediation and ongoing audit be allocated 60% to Defendants and 40% to Lumos, consistent with the cost-sharing arrangement established in the Court's March 8, 2023 stipulation and order appointing Brindoval as the neutral forensic examiner. Brindoval's total fees to date are $218,600. Under the existing 60/40 allocation, $131,160 is borne by Defendants/NovaSpark and $87,440 is borne by Lumos. Future forensic fees associated with the remediation and audit process should follow the same allocation, with Defendants bearing the majority share.
6. Mutual Non-Disparagement. Both sides should agree to a mutual non-disparagement provision prohibiting public statements that are derogatory or critical of the other party in connection with the litigation or the underlying events. This protects both Lumos's and Defendants' reputations in the marketplace.
7. Confidentiality of Settlement Terms. The financial terms of the settlement and the terms of the personal covenants should be subject to a confidentiality provision. However, the stipulated permanent injunction should be a public court filing, as it provides ongoing protection for Lumos and serves as a deterrent visible to others in the industry. Lumos's ability to reference the existence of the injunction — without disclosing the monetary terms — is important for its credibility with institutional clients and potential employees.
V. Mediation Strategy
The mediation is scheduled for September 6, 2023 before the Honorable Eleanor Voss (Ret.) at Tidewater Mediation Group, located at 160 West Santa Clara Street, Suite 1050, San Jose, California 95113. Judge Voss is a well-respected mediator with extensive experience in complex commercial and intellectual property disputes, and she is known for her ability to evaluate cases critically and to communicate realistic assessments to the parties.
We recommend the following strategic approach to the mediation:
Mediation Brief. A comprehensive mediation brief should be prepared and submitted to Judge Voss no later than August 30, 2023 — one week in advance of the mediation session — in accordance with Tidewater's standard protocols. The brief should present the full damages analysis set forth in this memorandum, including the unjust enrichment range of $8.7 million to $12.4 million and the exemplary damages potential of $17.4 million to $24.8 million. The brief should be supported by excerpts from the Brindoval Preliminary Forensic Report, the deposition transcripts (particularly Yun's admission), and Mr. Kapoor's technical assessment. We recommend submitting the brief on a confidential basis to the mediator only, with a redacted version available for the joint session if appropriate.
Opening Session. In the opening joint session, we recommend presenting the unjust enrichment analysis and the exemplary damages multiplier in a clear, direct manner calculated to maximize Lumos's leverage. The presentation should emphasize the strength of the forensic evidence, the damaging nature of Yun's deposition admission, the coordinated nature of the misconduct, and the Court's preliminary injunction ruling as an indicator of likely success on the merits. The goal of the opening session is to establish in the minds of Defendants and their counsel that Lumos's case is strong, that the damages exposure is substantial, and that the litigation risk for Defendants is significant.
Opening Demand. We recommend that Lumos open with a demand of $7.5 million in monetary compensation, plus all non-monetary terms described in Section IV.D above. This opening demand is within the unjust enrichment range and provides sufficient room to negotiate downward toward the target settlement range of $4 million to $6 million without signaling an immediate willingness to accept a figure at the lower end. The gap between the opening demand and the target range also allows for a structured negotiation process that respects the mediator's role and gives Defendants the sense that meaningful concessions are being made.
Key Arguments in Private Caucus. During private caucus sessions with Judge Voss, we should emphasize the following points:
• Collection risk. This is perhaps the most powerful argument for why Defendants should settle now. NovaSpark's cash is running out. Every month that passes without a settlement reduces the amount of money available to satisfy a judgment. If NovaSpark exhausts its funds before trial, Lumos will be left with a judgment against a shell entity and three individual defendants with limited personal assets. Settling now, while NovaSpark still has sufficient funds to make a meaningful payment, is in everyone's interest.
• Yun's deposition admission. This is a "smoking gun" that effectively eliminates the independent-development defense. We should ensure that Judge Voss understands the significance of Yun's admission and communicates its implications to Defendants' counsel during private caucuses. Defense counsel will have difficulty arguing that NovaSpark developed its technology independently when one of the three co-founders has admitted under oath to using Lumos's notebooks in NovaSpark's product development.
• Exemplary damages exposure. The threat of willful-and-malicious findings and a doubling of compensatory damages should be emphasized as a significant litigation risk for Defendants. Even if Defendants believe they can reduce the compensatory damages figure at trial, the exemplary damages multiplier creates an outsized exposure.
• Preliminary injunction ruling. The Court's February 17, 2023 order granting the preliminary injunction signals the Court's view on the merits and should be cited as evidence that the case is not a close call on liability.
Non-Monetary Terms. We should communicate clearly and early in the mediation that the non-monetary terms — stipulated permanent injunction, forensic remediation, personal covenants, and audit rights — are non-negotiable for Lumos. These terms are essential to achieving Lumos's core business objective of protecting its proprietary technology. We anticipate that Defendants may be more willing to accept these non-monetary terms than to agree to a larger monetary payment, which makes them a potentially useful point of leverage in the negotiation.
Mediation Fee. Tidewater Mediation Group charges a fee of $18,500 for the full-day mediation session, which is to be split equally between the parties ($9,250 per side).
VI. Fees and Costs Summary to Date
The following table summarizes the fees and costs incurred in connection with this matter through the date of this memorandum, as well as the projected allocation:
For reference, the applicable Calderwood & Harkness billing rates for this matter are as follows:
• Marcus Delgado, Responsible Partner: $1,150 per hour
• Catherine Bloom, Billing Partner: $1,050 per hour
• Wendy Schultz, Lead Associate: $725 per hour
• Omar Siddiqui, Working Associate: $575 per hour
Estimated Additional C&H Fees Through Trial (If No Settlement). If the mediation does not result in a settlement and the case proceeds through trial, we estimate that Calderwood & Harkness will incur an additional $800,000 to $1,200,000 in legal fees, depending on the complexity and duration of expert discovery, dispositive motion practice, and trial preparation. This estimate does not include additional Brindoval forensic costs for supplemental analysis, the retention of a damages expert for trial testimony, or potential rebuttal expert costs. When these additional expert costs are factored in, the total additional cost of proceeding through trial could reach $1.2 million to $1.6 million.
These projected costs further support the recommendation to seek a negotiated settlement at mediation, as the cost savings from avoiding trial enhance the effective net recovery to Lumos from a settlement in the recommended $4 million to $6 million range.
VII. Recommendation
Based on the analysis set forth in this memorandum, we recommend the following course of action:
1. Proceed with Mediation. Lumos should proceed with the September 6, 2023 mediation before the Honorable Eleanor Voss (Ret.) at Tidewater Mediation Group, seeking a settlement in the monetary range of $4 million to $6 million, together with all non-monetary terms identified in Section IV.D of this memorandum — specifically: a stipulated permanent injunction, comprehensive forensic remediation supervised by Brindoval, five-year personal covenants by each individual defendant, 18-month audit rights, mutual non-disparagement, and confidentiality of financial settlement terms.
2. Prepare and Submit Mediation Brief. A mediation brief presenting the full damages analysis and evidentiary record should be prepared and submitted to Judge Voss no later than August 30, 2023 (one week in advance of the mediation). We will circulate a draft of the mediation brief for partner review by August 23, 2023.
3. If Mediation Is Unsuccessful, Proceed to Trial. In the event that the mediation does not result in a settlement on acceptable terms, we recommend proceeding to trial. The evidentiary record supports a substantially larger recovery than the recommended settlement range, and the exemplary damages exposure provides significant additional motivation for Defendants to settle. If the mediation produces no resolution, we should promptly move to complete expert discovery and prepare for trial.
4. Client and Board Briefing. In advance of the mediation, we recommend scheduling a meeting with Lumos CEO Priya Chandrasekaran and General Counsel Thomas Brightwell to discuss the settlement parameters and obtain authorization to negotiate within the recommended range. We further recommend briefing Lumos's board representatives — Helen Trask (Ridgeline Ventures) and Jonathan Holt (Northstar Growth Equity) — on the mediation strategy and settlement range, to the extent authorized by management and consistent with board governance protocols.
VIII. Conclusion
The evidence developed through discovery to date — including the Brindoval forensic findings, the deposition testimony of each individual defendant, and the technical assessment of Meridian's Naveen Kapoor — supports a substantial recovery for Lumos Analytics Inc. The unjust enrichment conferred upon NovaSpark through its use of Lumos's misappropriated trade secrets is estimated at $8.7 million to $12.4 million, and the potential for exemplary damages amplifies this exposure to $17.4 million to $24.8 million. This damages framework provides Lumos with strong negotiating leverage in the upcoming mediation.
The recommended settlement target of $4 million to $6 million, when combined with the robust non-monetary protections described herein — the stipulated permanent injunction, forensic remediation, personal covenants, and audit rights — represents a favorable outcome that appropriately balances Lumos's potential recovery against the litigation risks, collection uncertainty, cost savings, and speed of resolution. Most importantly, a negotiated settlement with comprehensive injunctive and remediation terms achieves Lumos's primary business objective: the complete and verified elimination of its proprietary technology from NovaSpark's systems and the individual defendants' possession, with ongoing safeguards to ensure compliance.
We are available to discuss this analysis at your convenience and will proceed with preparation of the mediation brief upon your approval. Please do not hesitate to reach out with any questions, comments, or requested revisions.
Respectfully submitted,
CALDERWOOD & HARKNESS LLP
By: ________
Wendy Schultz Lead Associate
By: ________
Omar Siddiqui Working Associate
Enclosures:
1. Brindoval Digital Forensics LLC — Preliminary Forensic Report Summary (May 1, 2023)
2. Meridian Cybersecurity Advisors Inc. — Technical Assessment of Development Time Savings (July 12, 2023)
3. Deposition Transcript Excerpts — Rajiv Anand (May 15, 2023), Serena Mallick (June 7, 2023), and Derek Yun (June 22, 2023)
PRIVILEGED AND CONFIDENTIAL — ATTORNEY WORK PRODUCT This memorandum is protected by the attorney-client privilege and the work-product doctrine. Unauthorized disclosure is strictly prohibited.
Damages Theory | Low Estimate | High Estimate
Unjust Enrichment (Primary Theory) | $8,700,000 | $12,400,000
Reasonable Royalty (Alternative) | $5,000,000 | $10,000,000
Exemplary Damages (2× Unjust Enrichment) | $17,400,000 | $24,800,000
Vendor / Firm | Total Fees | Allocation
Calderwood & Harkness LLP (through July 2023, approximate) | ~$1,000,000 (estimated; final total will be confirmed with final billing reconciliation and will likely exceed this figure) | Borne entirely by Lumos
Brindoval Digital Forensics LLC | $218,600 | 60% Defendants ($131,160) / 40% Lumos ($87,440) per Court order dated March 8, 2023
Meridian Cybersecurity Advisors Inc. | $94,300 | Borne entirely by Lumos
Tidewater Mediation Group (estimated) | $18,500 | Split equally ($9,250 per side)

PRIVILEGED AND CONFIDENTIAL — ATTORNEY WORK PRODUCT

CALDERWOOD & HARKNESS LLP INTERNAL MEMORANDUM

MATTER NUMBER: 1008-00005

CALDERWOOD & HARKNESS LLP

Record

Parties

Dates

Amounts