Ooak Data
Applied AI research lab that turns real company data into reinforcement-learning environments where AI agents learn real-world work. YC S26, bootstrapped, working with leading AI labs.
What is Ooak Data?
Ooak Data is an applied AI research lab that turns real company data into reinforcement-learning (RL) environments where AI agents learn to do real-world work. Its pipeline sources multimodal data directly from real companies (documents, communications, and tool usage) with the full organizational context preserved, runs it through an automated multimodal anonymization step that transforms names, dates, and proprietary content into "digital twins" while keeping the structure, relationships, and complexity intact, and then generates expert-level tasks calibrated on the latest frontier models: multi-step, multi-tool workflows designed to expose model weaknesses rather than confirm strengths. The output is training and evaluation data for frontier AI labs building agentic systems. Ooak Data is a Y Combinator S26 company and describes itself as bootstrapped and very early, with three co-founders plus one ML engineer, already working with leading AI labs and actively hiring (including for a Chief Data Officer). It was co-founded by Pierre-Louis Vouteau and Grégoire. Because it is a small, early-stage lab, treat headcount and scale as tiny and fast-moving, and expect a research-and-data-engineering culture close to the frontier-model ecosystem.
Mission & values
Turn the messy reality of real company data into rigorous reinforcement-learning environments, so AI agents can learn to do real-world work and be tested against it honestly.
Qualifications
Ooak Data hires close to the frontier-AI-data frontier: machine-learning and research engineers, data engineers (multimodal pipelines, anonymization, environment generation), applied researchers who design agentic RL tasks and evaluations, and senior data leadership (it has publicly sought a Chief Data Officer). Application flow lives at ooakdata.com and via its YC and LinkedIn pages. The strongest fit is people with hands-on experience building RL environments, agent evaluations, or data pipelines for LLM training, plus comfort working with messy real-world enterprise data and frontier-model tooling. As a bootstrapped early-stage team of a handful of people, hires are highly leveraged and equity is a meaningful part of compensation. Work is described as remote-friendly and globally distributed.
Tech stack
Leadership
Hiring process
-
1
Apply at ooakdata.com or via YC / LinkedIn
Browse open ML, research, and data roles and reach out through the site or its YC and LinkedIn pages.
-
2
Founder screen
For a team this small, early conversations happen directly with a co-founder. Confirms fit, depth, and working style.
About 5 days
-
3
Technical deep-dive
Discussion and/or exercise on RL environments, agent evaluations, or multimodal data pipelines, matched to your area.
About 7 days
-
4
Work session or paid trial
Common at YC-stage labs: a paid trial or working session on a real problem before either side commits.
About 7 days
-
5
Offer with equity
Offer with base plus meaningful early equity given the bootstrapped, early stage.
About 3 days
Funding
Awards & recognition
-
Y Combinator S26 batch · 2026
Y Combinator
-
Working with leading AI labs on RL environments · 2026
Ooak Data
Company information
- Industry: AI Training Data / RL Environments / Applied AI Research
- Location: Remote-First (Global)
- Founded: 2025
- Company size: 1-10
- Website: ooakdata.com ↗
- LinkedIn: Company page ↗