Team / The people
We bring together research, evaluation, and the production of challenging training data in one founding team.
01 / CEO
Visiting researcher at Stanford, with a master’s degree from Fudan University. Advised by Jiajun Wu and Diyi Yang.
Research in agents and multimodal evaluation, including MetaGPT, AFlow, and Video-Bench.
5,900 citations · 80,000 GitHub stars
02 / Algorithms Lead
PhD student in Stanford’s Computer Science department and AI Lab, focused on evaluation and training-data pipelines.
Open-source contributor to Terminal-Bench, OpenThoughts, and OpenThoughts-Agent.
03 / Benchmark & Data Lead
Graduate student in Computer Science at Stanford, with experience at NVIDIA and the University of Michigan.
Co-author of Harbor-Index and creator of SWE-ZERO-12M, SWE-smith Multilingual, and Agents’ 100m. Works with Harbor, a framework for evaluating and optimizing agents in sandboxed task environments.
Join us in building datasets, tools, and evaluation systems for capable AI.