OOAK Lab
Our research lab studies how real work gets done, so AI agents can learn to do it well.
Research & Blog
Our writing on evaluating and training AI agents with real-world work.
Latest How we protect a company's history when training AI · 5 min read · Grégoire Lamy
When a company licenses its history to train AI, the first question is about trust, not price. Here's exactly what happens to the data, and why we treat anonymization as a matter of detail. Read the essay Why synthetic benchmarks are holding AI back · 10 min read · Pierre-Louis Vouteau
The AI industry over-indexes on synthetic benchmarks that don't reflect real-world complexity. Models that ace leaderboards routinely fail in production. Read the essay The data problem in agentic AI · 11 min read · Pierre-Louis Vouteau
AI agents need fundamentally different data than chatbots: multi-step interaction trajectories, tool-use demonstrations, and multi-modal workflow data. This data barely exists. Read the essay From sandbox to production: the gap nobody talks about · 12 min read · Pierre-Louis Vouteau
80% of AI projects fail in production. The problem is not the models. It's the gap between evaluation conditions and deployment conditions. Read the essay News
Announcements, press coverage and events from the OOAK team.
Want to co-author a paper with us?
We collaborate with researchers and practitioners on papers, benchmarks and in-depth articles about AI agents and real-world work.
Get in touch with our team at contact@ooakdata.com
Careers
Help us shape the next generation of AI.
A fast-growing team across Paris and San Francisco.