3:26-cv-09634
Databricks Inc v. Nom Nom Data Inc
I. Executive Summary and Procedural Information
- Parties & Counsel:
- Plaintiff: Databricks, Inc. (Delaware)
- Defendant: Nom Nom Data Inc. (Delaware) and Nom Nom AI Inc. (British Columbia, Canada)
- Plaintiff's Counsel: Fenwick & West LLP
- Case Identification: 3:26-cv-09634, N.D. Cal., 09/04/2026
- Venue Allegations: Venue is asserted in the Northern District of California because a substantial part of the events giving rise to the claim occurred in the district. Specifically, Defendants allegedly directed numerous communications, including infringement allegations and licensing demands, to Plaintiff Databricks, which is headquartered in the district.
- Core Dispute: Plaintiff Databricks, Inc. seeks a declaratory judgment that its data platform does not infringe Defendants' patent related to the automated management and optimization of Extract, Transform, and Load (ETL) data processes.
- Technical Context: The technology concerns systems for monitoring and automatically optimizing large-scale data processing pipelines, a critical function for efficiency and performance in modern cloud-based data analytics and business intelligence platforms.
- Key Procedural History: This declaratory judgment action was filed by Databricks following approximately nine months of escalating communications from Nom Nom. These communications allegedly included unsolicited acquisition pitches, multiple sets of element-by-element claim charts accusing Databricks of infringement, threats of contingency-fee litigation, and licensing demands targeting a claimed "$38.7 billion per year" market exposure.
Case Timeline
| Date | Event |
|---|---|
| 2024-02-21 | '995 Patent Priority Date |
| 2025-05-13 | '995 Patent Issue Date |
| 2025-10-21 | Nom Nom sends initial "M&A discussion" pitch to Databricks' CEO |
| 2025-11-05 | Nom Nom sends follow-up email |
| 2025-12-09 | Nom Nom sends email with a claim chart alleging infringement |
| 2026-01-15 | Nom Nom identifies its outside counsel to Databricks |
| 2026-02-17 | Nom Nom sends a "Competitive Licensing Analysis" and corrected claim charts |
| 2026-07-06 | Nom Nom introduces newly retained litigation counsel, the Devlin Law Firm |
| 2026-07-15 | Counsel for both parties hold a call; Nom Nom sends an updated claim chart |
| 2026-08-19 | Nom Nom holds another call with Databricks' counsel, presenting licensing options |
| 2026-09-04 | Complaint for Declaratory Judgment filed |
II. Technology and Patent(s)-in-Suit Analysis
- Patent Identification: U.S. Patent No. 12,298,995, "Systems, Methods, and Computer-Readable Media for Managing an Extract, Transform, and Load Process," issued May 13, 2025 (the "'995 Patent").
The Invention Explained
- Problem Addressed: The patent's background section describes that pre-defined Extract, Transform, and Load (ETL) processes, which are used to combine data from multiple sources, often suffer from performance issues. These problems arise from mismatches between the designed process and the actual hardware resources (e.g., memory, processor capacity) or from unforeseen changes in business requirements after the initial design '995 Patent, col. 1:31-44
- The Patented Solution: The invention is a system that monitors an ETL process during execution by receiving "ETL event data" '995 Patent, col. 2:53-55 It processes this data to identify a specific task or resource that is limiting the overall performance '995 Patent, col. 2:55-61 Upon identifying a bottleneck, the system "automatically" modifies the process to improve performance, for example by splitting a problematic task into smaller subtasks and rescheduling them '995 Patent, abstract '995 Patent, col. 2:46-52 This creates a dynamic feedback loop for real-time optimization.
- Technical Importance: The technology aims to provide real-time, automated error handling and performance optimization for complex data pipelines, a significant improvement over static workflows that require manual intervention to diagnose and fix bottlenecks '995 Patent, col. 4:57-66
Key Claims at a Glance
- The complaint seeks a declaration of non-infringement of claims 1-14 and 19 ('995 Patent, ¶34). The core technology is captured in independent claims 1, 10, and 19.
- Independent Claim 1 (System Claim) includes these essential elements:
- receiving ETL event data associated with an execution of a set of ETL tasks;
- processing the ETL event data to identify a first ETL task of the set of ETL tasks limiting a performance of the ETL process;
- processing the ETL event data to identify a resource limiting the performance of the ETL process;
- splitting the first ETL task into two or more subtasks; and
- automatically modifying a set of task instructions associated with the first ETL task to dynamically schedule the two or more subtasks as part of the ETL process to improve performance.
III. The Accused Instrumentality
Product Identification
The "Databricks Data Intelligence Platform," which the complaint describes as an "integrated cloud-based computing system" comprising numerous cooperating services rather than any single product in isolation Compl. ¶29 Compl. ¶35
Functionality and Market Context
The accused instrumentality is a comprehensive data analytics and AI platform. Nom Nom's infringement allegations, as recited in the complaint, target a wide array of its features, including Lakeflow, AI/BI Genie, Databricks SQL, Mosaic AI, Delta Live Tables, and Unity Catalog, among others Compl. ¶35 The complaint alleges the infringement theory is based on the collective operation of these integrated components Compl. ¶35 Databricks is positioned in the complaint as a highly innovative and major player in the data analytics industry (Compl. ¶¶18; Compl. ¶23).
No probative visual evidence provided in complaint.
IV. Analysis of Infringement Allegations
The complaint for declaratory judgment outlines Databricks' arguments for why its platform does not infringe the '995 Patent. The following table summarizes these non-infringement contentions as they relate to the elements of Claim 1.
- '995 Patent Infringement Allegations
| Claim Element (from Independent Claim 1) | Alleged Infringing Functionality | Complaint Citation | Patent Citation |
|---|---|---|---|
| process the ETL event data to identify a first ETL task of the set of ETL tasks limiting a performance of the ETL process... | The complaint asserts the accused platform does not perform this step, stating it has "no way of determining whether any ETL task is limiting performance." | ¶37 | col. 2:55-58 |
| split the first ETL task into two or more subtasks | The complaint asserts the accused platform does not perform this step, as there is "no way to split an existing task that is part of an executing ETL process." | ¶38 | col. 2:46-52 |
| automatically modify a set of task instructions associated with the first ETL task to dynamically schedule two or more subtasks as part of the ETL process to improve the performance of the ETL process | The complaint asserts the accused platform does not perform this step, as there is "no way to modify task instructions or schedule new subtasks as part of an executing ETL process." | ¶38 | col. 2:58-61 |
- Identified Points of Contention:
- Technical Questions: The central dispute appears to be factual and technical: does the Databricks platform actually perform the specific, dynamic, closed-loop optimization sequence described in the patent? The complaint frames this as a fundamental operational mismatch, alleging that the accused platform lacks the core capability to identify a performance-limiting task during execution and then automatically modify that specific task by splitting and rescheduling it Compl. ¶¶37-38 The court will likely need to examine detailed evidence of how Databricks' optimization and workload management features actually function.
- Scope Questions: The dispute raises the question of whether the claimed "automatic modification" requires rewriting or splitting an in-flight task, as Databricks' arguments suggest Compl. ¶38, or if it could be interpreted more broadly to cover other automated resource allocation and job scheduling adjustments that the Databricks platform may perform.
V. Key Claim Terms for Construction
The Term: "identify a first ETL task... limiting a performance of the ETL process"
- Context and Importance: This term is critical because it defines the trigger for the patented invention's corrective action. Databricks' primary non-infringement argument is that its platform has "no way of determining" this Compl. ¶37 The case will depend on whether the general performance monitoring in the accused platform can be considered to "identify" a bottleneck task in the specific manner required by the claim.
- Intrinsic Evidence for Interpretation:
- Evidence for a Broader Interpretation: A party could argue that any system that monitors task execution logs and flags a slow-running task meets the ordinary meaning of "identify." The patent mentions processing execution logs, which is a common monitoring function '995 Patent, col. 10:25-30
- Evidence for a Narrower Interpretation: The patent links this identification to a specific subsequent action: modifying that task '995 Patent, claim 1 This suggests "identify" means more than just logging; it implies a definitive diagnosis that directly causes a specific modification to the identified task, a functionality Databricks claims it lacks.
The Term: "automatically modify a set of task instructions... to dynamically schedule"
- Context and Importance: This term defines the core inventive action. Practitioners may focus on this term because Databricks argues its platform has "no way to modify task instructions or schedule new subtasks as part of an executing ETL process" Compl. ¶38 The dispute will center on whether Databricks' workload management and optimization features constitute "automatic modification" as claimed.
- Intrinsic Evidence for Interpretation:
- Evidence for a Broader Interpretation: The term "modify" could be argued to cover a range of automated actions, such as altering resource allocations or re-prioritizing jobs in a queue, which are common in modern data platforms. The specification mentions modifying hardware configurations as one possibility '995 Patent, col. 12:43-50
- Evidence for a Narrower Interpretation: Claim 1 itself provides a narrower definition by tying the modification to specific actions: "split the first ETL task into two or more subtasks" and "dynamically schedule the two or more subtasks." This language suggests a specific type of modification-task decomposition and rescheduling-rather than just any automated system adjustment.
VI. Other Allegations
This is a declaratory judgment action for non-infringement, and as such, the complaint does not contain allegations of indirect or willful infringement against Databricks.
VII. Analyst's Conclusion: Key Questions for the Case
A central evidentiary issue will be one of technical operation: Does the Databricks Data Intelligence Platform, as it actually functions, contain a mechanism that performs the specific closed-loop process of (a) identifying a discrete task as the performance bottleneck during an ETL process's execution and then (b) automatically modifying that specific task by splitting it into subtasks and rescheduling them? The complaint frames this as a fundamental technical disconnect between the patent's claims and the accused platform's capabilities.
The outcome will also depend on claim construction: How narrowly will the court define the terms "identify a...task limiting a performance" and "automatically modify"? The key question is whether these terms require the specific, dynamic, in-flight task-splitting and rescheduling process described in the patent's embodiment, or if they can be construed more broadly to encompass the general, system-wide optimization, resource allocation, and job management features common to modern data platforms, which Nom Nom will likely argue are equivalent.