Research

Currently:

• AI safety / AI interpretability research at MATS in Berkeley, with Team Shard (Alex Turner, Alex Cloud). Thinking about: verbalization and language

• DNA language models: Evo3 and Evo Sapiens with the Hsu Lab, Arc Institute. Thinking about: data scaling and human health

Previously:

• CAMBRIA (Cambridge Bootcamp for Research in Interpretability and Alignment) in May 2026, where I played with natural language autoencoders for our research capstone.

• Protein structure prediction/design models with MinGyu Choi in RGB. Tried to improve Boltz’s protein-modeling realism by adding back environmental water molecules to the x-ray crystallography training data.