Researching flexible language models.

I helped develop FlexOLMo, and now apply related ideas in federated learning setups to train on privacy-sensitive cancer data across institutions through the Cancer AI Alliance.

Previously, I worked at TrueMedia with Oren Etzioni on deepfake detection systems intended to help safeguard the 2024 U.S. election.

I earned my BS/MS in Computer Science at the University of Washington, where I worked with Akari Asai in the H2Lab under Hannaneh Hajishirzi.

Current interests: adaptive LLMs, continual learning, flexible and modular models, and agents.