WAI is a research organization formed by NLP / ML / System PhD students from the University of Washington. We work on agentic research across the full stack: data, training, and evaluation.
AI has advanced faster than the understanding of it. Knowledge of how agents are built — what data they learn from, how they are trained, how they are judged — is largely closed source. We think the science behind them should be open.
So we do three things. We do research on agents. We open-source what we build: data, code, and systems. And we host tutorials and seminars on what is happening at the frontier. Science is better when it is shared, and we publish write-ups and findings as we go.
Our story
WAI started on a walk. A few of us were talking about how the field was at a fork in the road: frontier labs publish less and less, and academia drifts further from where the systems are actually built. Meanwhile more people than ever want to understand and use them. The gap between what is known and what is shared keeps growing. At WAI, we bridge that gap by building AI agents and openly sharing how we do it.
Organizers
- Hamish Ivison
- Junjie Oscar Yin
- Rulin Shao
- Steven Gao
Advisors
Projects
TMax — an open recipe for state-of-the-art terminal agents.
Team: Hamish Ivison*, Junjie Oscar Yin*, Rulin Shao, Teng Xiao, Nathan Lambert, Hannaneh Hajishirzi
Get in touch
Interested in collaborating or contributing? Reach out at wai.org.research@gmail.com or follow us on Twitter @waiorg.