I'm a recent McGill graduate with a B.Sc. in Statistics and Computer Science, passionate about AI safety and alignment research, with a special interesting in AI Control, reward hacking, and scalable oversight. I've been involved in the effective altruism and AI safety communities, serving as co-president of EA McGill and founding AI Alignment McGill.
My background includes participating in the first iteration of the ML Safety Scholars program, leading an AI Safety Camp cyborgism project, and building tools like AlignmentSearch / Stampy Chat to help others learn about AI alignment. More recently, I've worked on AI Control research projects.
Some other small projects:
- PressBench: A benchmark measuring how much frontier models self-report they'd be able to benchpress if they were human.
- FriendBench: A benchmark evaluating the Friend-ness of frontier AI systems.
- n-d-chess: Exploring high-dimensional chess; analyzing how the number of squares each piece can attack scales based on the board size and dimension.
Outside of AI work, I enjoy chess, arm wrestling, and forecasting.
I'm starting a MSc at Mila in a few months.
Recent Work
- AlignmentSearch / Stampy Chat - Built a conversational AI agent for AI safety Q&A, now integrated into the Stampy.ai platform
- Feedback-based Control Protocols - Developed feedback-based AI control protocols for the Cambridge AI Safety Hub