From Agent Traces to Agent Simulations — Rustem Feyzkhanov, Snorkel AI
AI Engineer World's Fair 2026 · 20:24
AI data development and evaluation
Snorkel AI is a frontier AI data lab founded in 2019 out of the Stanford AI Lab. It develops specialized data, benchmarks, evaluation systems, and RL environments for frontier AI labs and enterprise teams, combining expert workflows with research-driven data development.
Snorkel’s $30 million Open Benchmarks Grants commitment supports researchers, domain experts, and open-source teams developing open benchmarks. Snorkel’s benchmark collaborations include Terminal-Bench, OSWorld 2.0, Agents’ Last Exam, and Senior SWE-Bench. The initiative also includes the Open Benchmarks Red Team, which helps benchmark builders identify and address weaknesses in their evaluations, and the Snorkel Research Fellowship, which supports independent evaluation research for PhDs and postdocs.
Snorkel also builds specialized agents for enterprise workflows, with engagements spanning use-case scoping, data and evaluation development, deployment, monitoring, and iteration.
AI Engineer World's Fair 2026 · 20:24
AI Engineer World's Fair 2026 · 17:15
AI Engineer Europe 2026 · 20:56
AI Engineer Europe 2026 · 20:40
AI Engineer World's Fair 2024 · 20:46
Loading more talks and speakers…
Affiliations reflect their AIE event listings and recorded appearances, not necessarily current employment.