I recently completed an MSc in computer science at EPFL, where my thesis studied how inductive reasoning develops during language model pre-training. I am now a research assistant with Prof. Ufuk Topcu at UT Austin.

My research focuses on reverse-engineering deep learning models: what internal mechanisms drive their behavior, how these emerge during training, and why they fail in unfamiliar settings. Currently, I study this in language models, especially how training data shapes what they learn and how they generalize. My goal is a predictive science of training that enables more reliable AI systems.