Technical co-founder of Datacurve and co-author of its DeepSWE benchmark. Former Google intern and Waterloo CS student focused on multimodal RL and browser-automation agents.
Background
Studied computer science at the University of Waterloo; interned at Google. YC lists research interests in multimodal reinforcement learning and browser-automation agents. Built UncleGPT, a planning tool, with Serena Ge before Datacurve.
What they run now
Appears to lead technical/research work: co-author of DeepSWE (May 2026), with its original tasks and handwritten verifiers.
Career
- 2024 – presentCo-founder, Datacurve
- —Intern, Google
- —University of Waterloo, Computer Science — Completion status unknown
Recent posts
3 posts archived · most engaged first, then the latest
Proud of all the progress made so far, but even more excited for what's next.
Our results for Opus 4.8 on DeepSWE are in!
Overall, better performance while being cheaper than Opus 4.7
Personal experience using these models has always seemed to not be aligned with existing benchmark numbers. We put a lot of thought into the analysis and in closing the gap.