WAI
WAI is a research organization formed by PhD students from the University of Washington working on open agentic research across the full stack—data, training, and evaluation.
At a Glance
- AI Researchers
- Developers
- Academia
AI Tools by WAI
(1)
TMax
Terminal Agent Training Framework
Discussions
No discussions yet
Be the first to start a discussion about WAI
Latest News
Products & Services
An open recipe for state-of-the-art terminal agents using outcome-only RL (GRPO).
A dataset of 14,600 RL environments built from a compositional pipeline with explicit control over difficulty and diversity.
A family of open-weights models (2B, 9B, 27B) trained using the TMax recipe.
Market Position
WAI fills the gap between frontier labs and academia by providing 'frontier-grade' open recipes for AI agents, competing with closed-source systems by offering open alternatives.
Leadership
Founders
Rulin Shao
PhD student at the University of Washington CS / NLP group. Researcher in Machine Learning and NLP with experience at Apple Machine Learning Research.
Junjie Oscar Yin
PhD student at the University of Washington focused on NLP research. Previous research at University of Washington.
Steven Gao
PhD student at the University of Washington.
Executive Team
Rulin Shao
Organizer
PhD student at University of Washington NLP group.
Junjie Oscar Yin
Organizer
PhD student at University of Washington.
Board of Directors
Founding Story
WAI was started by University of Washington PhD students to bridge the gap between closed-source frontier AI labs and academia by openly sharing the science and data behind AI agents.
Business Model
Revenue Model
Open Source / Academic Research supported by university resources.
Pricing Tiers
All research artifacts, datasets, and model weights are publicly available.
Target Markets
- AI Researchers
- Developers
- Academia
- Building terminal-based AI agents
- Researching agentic RL recipes
- Scaling open-weights agent models
- Allen Institute for AI