I work on mechanistic interpretability and evaluation of language models, currently as an independent researcher. My focus is on distinguishing what models retrieve from what they compose, and on building evaluation methods that hold up outside the lab.
Before this I spent a decade in clinical care and health research — pediatric critical care nursing, then human-computer interaction research at Stanford, where I earned an MS. I founded and exited Syminar, a clinical education platform. The throughline is making technical systems legible to the people who have to live inside them.
In progress. Notes on interpretability experiments will land here.