Snorkel AI

Snorkel AI's roots trace to the Stanford AI Lab, where PhD student Alex Ratner, professor Christopher Ré, and a group of Ré's students and postdocs (Braden Hancock, Henry Ehrenberg, Paroma Varma, Manas Joglekar, and Vincent Chen) began researching 'weak supervision' -- techniques to train machine learning models without large hand-labeled datasets, initially exploring applications like combating human trafficking. That research produced the open-source Snorkel framework and a widely cited 2017 VLDB paper ('Snorkel: Rapid Training Data Creation with Weak Supervision'), which introduced 'data programming': letting subject-matter experts encode labeling logic as heuristics/code that can be applied programmatically across massive datasets, rather than labeling examples one at a time. The team spun the research out as Snorkel AI in 2019, with Ratner as CEO and Ré (a 2015 MacArthur 'genius grant' recipient who also co-founded SambaNova Systems and the Apple-acquired startups Lattice Data and Inductiv) as a co-founder and board member.

Greylock -- via partner Saam Motamedi -- was Snorkel's earliest backer, co-leading a $3M seed round in mid-2019 and then leading a $12M Series A that October (with GV and In-Q-Tel participating); Motamedi joined Snorkel's board at that point and has remained involved through every subsequent round. The company's flagship product, Snorkel Flow, launched commercially in 2021 and became the vehicle for 'data-centric AI' -- the idea (popularized alongside figures like Andrew Ng) that iterating on training data, not just model architecture, is the highest-leverage way to improve ML system quality. Snorkel raised a $35M Series B in April 2021 (led by Lightspeed Venture Partners, joined by existing investors Greylock, GV, In-Q-Tel and Nepenthe Capital, plus new investors Walden International and BlackRock) alongside the launch of Application Studio. Four months later, in August 2021, it raised an $85M Series C co-led by BlackRock and Addition at a valuation of roughly $1 billion, making it one of the fastest Stanford spinouts to reach unicorn status; new investors Factory and Cooley joined existing backers Greylock, GV, Lightspeed, Nepenthe Capital, Walden International and Stone Bridge Ventures. By that point the company had raised $135M total and counted Google, Apple, Intel, IBM, Microsoft, Accenture, DARPA, Stanford Medicine and Uber among its customers/partners.

The launch of ChatGPT in late 2022 upended Snorkel's core market: large language models could now auto-label huge volumes of data cheaply, temporarily slowing enterprise demand for programmatic labeling tools as some customers paused pilots to explore LLM-based labeling directly. Snorkel responded by pivoting further up the value chain -- from raw data labeling toward fine-tuning, RAG data curation, and especially AI model evaluation, arguing that domain-expert-authored evaluation and 'post-training' data would become the critical bottleneck for enterprise and frontier AI alike. It signed and expanded partnerships and customer engagements with Google Cloud, Microsoft Azure, Snowflake, Aimpoint Digital, BNY, and others, and won government contracts including a U.S. Air Force data-labeling award (2023), Army xTech AI Grand Challenge recognition (2025) and a completed Defense Innovation Unit challenge (2025). In February 2024, QBE Ventures took an equity stake and partnered with Snorkel on generative-AI use cases in insurance; in May 2024 the company added four new C-suite executives (Chief Development, Marketing, Financial and Product Officers) as it scaled go-to-market.

Business model

B2B enterprise software plus expert-services/data-as-a-service hybrid: licenses its Snorkel platform (Snorkel Flow, Snorkel Evaluate) to enterprises and government agencies, and increasingly sells 'Data-as-a-Service' engagements where Snorkel's own staff and a large network of contracted subject-matter experts (doctors, lawyers, accountants, STEM PhDs, engineers) build custom datasets, evaluation suites, and RL/agentic environments for frontier AI labs and large enterprises.

Enterprise software subscriptions/licenses for its data-development platform, combined with services/data-as-a-service contracts (often large, multi-year enterprise or frontier-lab engagements) for custom dataset and benchmark creation; company has declined to disclose revenue publicly.