Callum McDougall
Bio
Callum McDougall is an Interpretability Researcher at Google DeepMind, based in London. He co-founded and ran the ARENA (Alignment Research Engineer Accelerator) program starting in 2022, a widely-used open-source curriculum for technical AI safety upskilling covering deep learning fundamentals, mechanistic interpretability, and reinforcement learning. He holds an MMath from the University of Cambridge (2018-2022), where he ranked 11th out of 222 and received the Dr J.A.J. Whelan Prize in Mathematics. Before joining DeepMind, he was a Member of Technical Staff at Anthropic (Feb-Aug 2024) working on interpretability, and prior to that worked at Jane Street. His research focuses on mechanistic interpretability and sparse autoencoders, including contributions to the Gemma Scope and Gemma Scope 2 projects, which release open SAEs for the Gemma model families to support safety and interpretability research.
Links
- Personal Website
- https://www.perfectlynormal.co.uk/
- Twitter / X
- LessWrong
- callummcdougall
Grants
No grants recorded.
Discussion
Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.
Details
- Last Updated
- Mar 22, 2026, 2:49 PM UTC
- Created
- Mar 20, 2026, 3:00 AM UTC