Kevin Han

I’m a third-year PhD student at Carnegie Mellon University advised by Niloofar Mireshghallah and Amir Farimani. My research is focused on learning for automated scientific discovery using LLMs, primarily focused on materials and drug discovery, as well as large-scale atomistic simulation.

Previously, I worked on unsupervised domain adaption for LLMs at JPMorgan Chase’s Commercial Bank AI team. I also worked at Lawrence Berkeley National Laboratory on machine learning interatomic potentials. I graduated from UC Berkeley with a Bachelor’s in Computer Science, where I was fortunate to work with Gerbrand Ceder and Bowen Deng on CHGNet where I wrote high performance C code for graph construction.

I come primarily from a machine learning background rather than a chemistry or biology one. My work spans PyTorch, SGLang, NeMo Gym, miles, Cython, and performance programming in C. I also enjoy full-stack engineering, and have built large-scale systems for startups and pedagogy software for the UC Berkeley EECS Department.

If you have questions about Berkeley, CMU, finding internships or research, or general life stuff, feel free to email me at kevinhan@cmu.edu.