Hello! I'm Ollie Jaffe.

I'm currently creating automated AI safety scientists.

Previously I worked at OpenAI for 2 years as a contractor researching dangerous capability evaluations, with a focus on AI R&D and sandbagging capability evaluations. Some of my previous research includes SWE-Bench Verified, PaperBench, and MLE-Bench, and a small sandbagging eval.

[LastName]ollie@gmail.com