Hi, I'm Prakrat!
I'm work on technical AI Safety research and study Math and Computer Science at UC Berkeley. I'm currently doing the Pivotal Fellowship under Jérémy Scheurer, where I'm working on evaluating sabotage propensity in coding agents. I'm also finishing up my project on evaluating self-preservation drives in LLMs. On campus, I'm a member of the Berkeley AI Safety Initiative and Machine Learning @ Berkeley. I also help teach the Intro to AI safety DeCal.
Previously
I interned as a Software Developer at Amazon where I worked on RAG applications for financial data, tried to make a consumer productivity app with my friend Ady, helped teach the biggest class at Berkeley, and organized a retreat for university AI Safety groups. See the full list of my projects here.
Interests
I enjoy reading, hiking, teaching, hearing about people's interests, learning new skills, playing ultimate frisbee, table tennis, spikeball, and board games, and listening to house music.