Alex completed Part III of the Mathematical Tripos at Cambridge, with a focus on algebraic number theory.
During his degree, he took part in machine learning research focused on mechanistic interpretability, including work on establishing formal guarantees about neural network behaviour and on understanding how early layers of LLMs process contextual information on typical inputs.
His current research interests include mechanistic interpretability, the dynamics of neural network training, and generative modelling. He is particularly interested in diffusion and flow-based models, sampling methods, and their connections to stochastic calculus and optimal transport.
Outside of research, Alex enjoys reading and playing chess.
