Since 2021, my interests have expanded to technical work: I studied machine learning, applied math, and programming to upskill in AI safety.

I was initially interested in large language model (LLM) evaluations. I led a project on machine learning for scientific reasoning where my team and I evaluated LLM capabilities relevant to scientific tasks. Our project report is available here.

More recently, I have been working through problems in the LLM counterpart of neuroscience, namely, mechanistic interpretability.

My GitHub page contains some of my experiments in technical AI safety research.