My Approach to AI in Mathematics
My work tries to make frontier models genuinely useful for research by combine the creative, pattern-matching capabilities of LLMs with the logical rigor of computation and verification.
Projects
-
Pachner-move simplification (exploratory)
Finding potential counter-examples for the 4-dimensional smooth Poincaré Conjecture—using deep reinforcement learning. The idea is to frame the problem as a game where an AI agent, built on Graph Neural Networks, learns a policy to simplify complex triangulations by choosing optimal sequences of Pachner moves.
-
Mini-lecture series: Peg Solitaire and basic AI ideas
Three short talks introducing convolutional nets, Q-learning, and transformers by way of a concrete game example.
→ Lecture 1 slides (PDF) -
VOA/MTC database
Curated tables for VOAs and modular tensor categories. A compact, checkable resource for routine lookups.
→ Access