Julius Broomfield

Hi, I’m Julius. I work on making language models robust, aligned, and epistemically responsible, and I’m a final year undergrad studying Computer Science and Philosophy at Georgia Tech 🐝, advised by Prof. Srijan Kumar.

I’m interested in the relationship between humans and AI systems, and how we make sure it goes well on both ends, for both the people interacting with these systems and the systems learning from them. I work on adversarial robustness [1], studying how safety guardrails fail in practice; pluralistic alignment [3] and preference representation [4], studying how models learn values and preferences and whose values they end up serving; and the epistemics effects of AI [5], thinking about how AI shapes what people know and believe.

Previously, I interned on the Copilot Infra team @ Microsoft AI, and as a Software Engineering intern on the Job Seeker Experience team @ LinkedIn. I’ve also been fortunate to work with mentors who have helped me grow, including Kellin Pelrine, Kartik Sharma, and Isaac Chepkwony.

In my free time, I like playing the drums, looking for new music, grappling, playing chess, and reading philosophy. I’m always happy to chat about research, collaborations, or anything else — please feel free to reach out!