Co-founder
Snorkel AI
Snorkel AI partners with frontier AI teams to build research-grade datasets, evaluation systems, and runnable environments that close specific model performance gaps where generic coverage runs out. The company develops bespoke data series—including expert demonstrations, reasoning traces, preference labels, and verifiable rubrics—alongside custom environments like repo and CLI tools, browser harnesses, and simulated stateful workflows. Every dataset and benchmark is co-developed and peer-reviewed with leading academic teams and frontier labs.
- Artificial Intelligence
- Developer Tools
- Software Only
- AI Agents
- Enterprise Software
Updated 5 days ago
Funding
Raised to date
$585MRaised to date based on public sources. This may differ from the amount the company actually raised and is based only on what is publicly available on the internet.
$488.3M across 4 disclosed rounds
+15$350M · Series E
Announced Sep 2026
Disclosed cumulative: $488.3M
4/4
Reported total $585M
Not on chart (date unknown): Series C Not disclosed
Incomplete round history. $96.8M of the $585M reported total isn’t matched to a known round (1 known round has an undisclosed amount).
RoundDateAmountValuationInvestors
Founders
Product
Problem
Frontier AI teams often hit performance ceilings when generic, off-the-shelf datasets and benchmarks fail to cover the specialized failure surfaces of their models. Building bespoke evaluation systems and high-quality training data in-house is time-consuming, expensive, and requires deep domain expertise that many teams lack.
Solution
Snorkel AI partners with frontier AI teams to build research-grade datasets, evaluation systems, and runnable environments tailored to exact model weaknesses. The company develops custom data series—including expert demonstrations, reasoning traces, preference labels, and verifiable rubrics—that ground models in specialized workflows. It also creates standard and custom environments such as repo and CLI tools, browser harnesses, and simulated stateful workflows that integrate with a client's existing tools and codebase. Every dataset and benchmark is the output of active research co-developed and peer-reviewed with leading academic teams and frontier labs, ensuring scientific rigor and practical utility.
Target Audience
Primary customers are frontier AI research teams and commercial labs that need custom datasets, evaluation systems, and runnable environments to close specific model performance gaps in specialized domains.
Features
- Bespoke dataset development covering expert demonstrations, reasoning traces, SME Q&A rationales, and tool-use demos
- Preference labeling and ranking services for patch/draft/report quality, trajectory QA, risk/safety calibration, and helpful/harmless scoring
- Rubric and verifiable outcome design including unit tests, deterministic graders, citation correctness checks, and numerical consistency scoring
- Custom environment construction spanning repo and CLI tools, browser/GUI harnesses, multi-step stateful workflows, and simulated environments
- Specialized agent development grounded in expert data for high-consequence decision workflows rather than generic copilots
- Research outputs co-developed and peer-reviewed with academic teams and frontier labs
- Public leaderboards such as SnorkelLegal, which evaluates agents within a simulated mid-market law firm across 200 frontier tasks with 15 simulated personas