Snorkel AI
Expert data development for specialized AI
CORE INFO
Snorkel AI develops expert-built training and evaluation data, benchmarks, and simulated environments for AI labs and enterprises building specialized models and agents. Its work covers domains such as finance, healthcare, insurance, government, legal work, and software engineering.
- Series E financing totaled $350 million at a $3.5 billion valuation on September 22, 2026, co-led by Insight Partners and S32.techcrunch.com
- Annualized revenue reached $375 million as of September 2026, with TechCrunch reporting 18× growth over the prior 12 months.techcrunch.com
- TechCrunch reported that Snorkel AI has shifted from programmatic labeling software toward completed datasets, synthetic data, and simulated environments for AI agents.techcrunch.com
WHY WE WOULD WORK AT SNORKEL AI
Advance Safer AI
Help define and advance the frontier of AI through expert data development. Your work can improve how specialized systems are trained, evaluated, and deployed in high-consequence fields.
Work on Frontier Research
Build on Stanford AI Lab roots alongside a research-driven team with 250+ peer-reviewed publications and recognition from NeurIPS, ICML, ICLR, and other leading conferences.
Build Agent Intelligence
Develop evaluation datasets, deterministic graders, and environments for browser, GUI, repository, command-line, and enterprise agents. Snorkel's work includes benchmarks such as Terminal-Bench and OSWorld 2.0.
Grow With a Scaling Company
Snorkel has expanded from programmatic labeling software into expert data, synthetic data, evaluation, benchmarks, and specialized agents. Its reported annualized revenue reached $350M–$375M in 2026, up from roughly $20M a year earlier.
Collaborate With Experts
Work across research, engineering, product, data operations, and domain specialists—including professors, lawyers, accountants, and other subject-matter experts—to solve difficult AI problems.
See Measurable Customer Impact
Help power production systems for organizations including Wayfair, Experian, Chubb, BNY, and the U.S. Air Force. Reported results include 99% accuracy, 10× faster development cycles, and 35% automation of Experian customer email responses.
MARKET AND TRACTION
GROWTH TACTICS
NOTABLE CUSTOMERS
KEY METRICS
✦ KEY METRICCOMPETITIVE ADVANTAGE
✦ KEY METRICMARKET POSITION
PRODUCT AND TECH
Expert Data Development
Creates specialized training, post-training, preference, and evaluation datasets using expert demonstrations, synthetic data, programmatic methods, rubrics, human review, and provenance tracking. This helps organizations address domain-specific model failures and improve AI reliability.
Snorkel Flow Platform
Snorkel's data-centric AI platform supports programmatic labeling, weak supervision, data curation, model development, error analysis, and human-in-the-loop workflows. It enables teams to build training data without relying solely on manual annotation.
Evaluation and Benchmarks
Develops evaluation datasets, deterministic graders, rubrics, leaderboards, and benchmarks such as Terminal-Bench, Senior SWE-bench, and OSWorld 2.0. These tools measure model and agent performance against task-specific and verifiable criteria.
Agent Evaluation Environments
Provides browser, GUI, repository, command-line, and simulated enterprise environments for testing multi-step, stateful, and tool-using AI agents. Environments can incorporate customer-specific tools, codebases, permissions, corpora, and workflows.
Specialized AI Agents
Builds custom agents for high-consequence workflows in areas including finance, insurance underwriting, healthcare, government, legal work, and software engineering. These systems are evaluated with domain-specific, programmatic pass/fail criteria rather than generic benchmarks.