Mathematician interested in understanding how AI systems work.
PhD in complex algebraic geometry from the University of Waterloo, Ontario, Canada. I build and fine-tune machine learning models, and I'm interested in research on what trained models learn and how they learn it, especially for AI safety.
Supervised and unsupervised learning, deep learning in PyTorch, LLM fine-tuning.
Python and SQL, feature engineering, model evaluation.
LLM agents with LangChain and LangGraph, and expert evaluation of model reasoning in math.
Algebraic geometry, moduli, stability, sheaf cohomology.
Machine learning projects, from on-device speech models to graph neural networks, and products I have built.
Open →ResearchBundle Networks for exploring fraud models, and my PhD and master's theses in algebraic geometry.
Open →BlogConjectures, sketches and half-finished thoughts on math and AI.
Open →