william [dot] computer

Hello. My name is William Sepesi.

I'm a machine learning researcher and engineer. I graduated from Washington University with a double major in computer science and mathematics.

My interests lie across reinforcement learning (particularly incomplete information), LLMs (particularly multimodal and/or local models), and mechanistic interpretability. I've worked as an ML engineer and researcher at Microsoft, Square, and a handful of startups in these areas and beyond. I was also the Head Teaching Assistant for a variety of ML courses during my undergrad.

Outside of ML I enjoy music, travel, biking, contemporary art, and speaking French. You can contact me at hello [at] william [dot] computer or [at] robot__fan on Twitter, or find me on Github [at] wsepesi.