We’re excited to announce March Capital’s investment in Snorkel AI’s $350 million Series E.
Earlier this year, we wrote about a shift underway in AI: agents take on increasingly complex work, the unit of progress is moving beyond tokens generated toward useful work delivered. What matters is: 1) can an AI agent complete the task, 2) at what cost, and 3) with what reliability.
That shift creates a new set of research problems.
An agent may operate for hours, use dozens of tools, navigate software, recover from mistakes, and produce business collateral. Advancing these systems requires environments in which they can learn, calibrate tasks to the edge of their capabilities, and verify failure points. This is the substrate for reinforcement learning.
Anyone chasing better model performance knows it’s a constantly moving target. Once a model masters one benchmark, researchers need a harder one. As agents take on longer and more consequential work, the environments, rubrics, and training data required to improve them grow more complex. When models improve, the data required to improve them becomes even more challenging to produce.
We first met Alex Ratner in 2022 and he joined us at the Montgomery Summit to present Snorkel the following year. By then, Alex and the Snorkel team had spent years building the research foundation for this moment. Born out of Chris Ré’s AI Lab at Stanford, the founding team helped pioneer data-centric AI, with more than 250 peer-reviewed papers spanning weak supervision, data curation, synthetic data, evaluation, and model training.
Today, that research powers Snorkel’s Agentic Data Factory. Snorkel works directly with research teams from world-leading labs, AI-natives and Fortune 500 enterprises to define, measure and evaluate AI systems. It produces and designs the benchmarks, RL environments, and training data required to move the frontier forward.
Its Open Benchmark Grants program with AI-natives like Harvey extends that work into the broader research community, openly sharing new benchmarks so researchers can independently probe emerging capabilities and failure modes, reproduce results, and collectively raise the bar for AI safety.
The market is already moving quickly in Snorkel’s direction. Since launching its expert Data-as-a-Service offering in September 2025, the company has achieved extraordinary growth and is working with top organizations like OpenAI, Anthropic, Google DeepMind, Thinking Machine Labs, Bank of America, and others on some of their hardest training and evaluation problems.
As the surface area of work AI can perform expands, so does the surface area of data required to teach and evaluate it. We believe Snorkel is building the data layer for that future, and we’re thrilled to partner with Alex and the Snorkel team in their journey.
article written by Solomon Hailu and Maya Matthews
Learn more about Snorkel in our MONTY podcast featuring Alex Ratner, Co-Founder & CEO of Snorkel AI: Beyond the Model: Building the Data Layer for the Next Era of AI | Alex Ratner, Snorkel AI
Snorkel Press Release, featured in Reuters: EXCLUSIVE: Snorkel AI valued at $3.5 billion amid surging demand for complex AI training data | Reuters