Building the Frontier Lab for Agentic Data

Introduction

Today, I’m incredibly excited to announce Snorkel’s $[350]M Series E financing [and employee tender] at a $3.5B valuation, co-led by Insight Partners and S32, with significant participation from existing investor Addition. The round included new investors March Capital, Blumberg Capital, Allegis Capital, Frontline, Standard, One Prime Capital, Third Point Ventures, and D.E. Shaw Ventures, along with existing investors Greylock, Lightspeed, GV, Factory, BlackRock, Prosperity7, and Wells Fargo.

Over the last year, our technology and research-centric approach to AI data development has driven [18x+] YoY growth and helped us cross a $[350M] annualized revenue run rate in just twelve months. We’re grateful to the customers and partners who enabled this, and excited to invest this new funding in supporting their continued success.

Snorkel started as a research project a decade ago at Stanford, founded on the idea that AI would inevitably become “data-centric”, and that data deserved to be the centerpoint of proper academic study.

Snorkel then spun out as a company [X] years ago, founded on the idea that AI data should be approached as a software technology problem – not just a staffing or crowdsourcing one.

Today, this research- and software-centric approach to AI data is more important than ever before. AI data needs have gone through a fundamental phase shift – from simpler, volume-centric “Data 1.0” workloads, to highly complex agentic “Data 2.0” environments and datasets – and this new regime of data requires a new type of data company: a frontier data lab that blends research, platform technology, and expert humans to push the frontiers of AI capabilities.

We’re excited to keep building our vision of a true frontier data lab with the new round of funding.

In particular, we’re going to be investing in several core areas that we believe are critical to safely and effectively advancing AI, building on our decade of R&D in these areas:

1. Building the agentic data development platform: As AI capabilities continue to advance, we believe that frontier data can only be driven by a self-reinforcing flywheel of human experts accelerated and improved by AI, and AI models supervised and refined by human expert input.
2. Advancing open benchmarks: We believe that developing an increasingly rich surface area of diverse, continuously adapted, and open benchmarks is the most robust way to measure and advance AI.
3. Driving a diversified spectrum of intelligence: We believe that there will be a rich gradient from general models to highly specialized agents, each evaluated and trained on their own unique datasets and environments that need to be continuously created, curated, and refined.
4. Data as the final frontier of AI research: We believe that as model and algorithmic approaches continue to standardize, data becomes the most interesting area of AI research.

If any of these ideas excite you, please shoot us a note!  We’re hiring 🙂