About
I’m an undergraduate at Penn State — computer science with minors in physics and AI, Schreyer Honors College — working toward a career in mechanistic interpretability research. I came up through physics: my first real project was a Monte Carlo radiative-transfer code for neutron-star atmospheres, the kind of work where nothing counts until it matches an exact answer. Transformers get the same treatment here: I trained a small character-level model and traced a half-built copying mechanism in its weights, read GPT-2’s IOI circuit through three methods to see where each one goes blind, and am currently testing whether Anthropic’s global-workspace result replicates from scratch. Everything on this site is built or measured by me — controls included, especially the ones that broke my own hypotheses.