Aarush Sah

I work on AI. For the past few years that has mostly meant evaluation — deciding what to measure, and building the infrastructure to measure it properly. Outside of work I read a lot: fantasy and science fiction mostly, the longer the better.

If you've heard of me, it's probably because of openbench, an open-source framework I built at Groq for running language-model benchmarks reproducibly.