Research Interests
I study the mathematical and algorithmic foundations of machine learning and AI, drawing on optimal transport, information theory, statistics, and optimization. My current interests include:
Optimal transport theory and the Gromov-Wasserstein (GW) problem: Geometry, statistics, algorithms, and alignment of heterogeneous data.
AI mechanistic interpretability, safety, and alignment: Principled approaches for causal abstractions, information-flow analysis, and steering of neural networks, especially LLMs.
Information theory (including quantum information theory): Learning, privacy, security, and information-theoretic foundations of modern AI.
Robust statistics and decision-making: Inference and learning under local (perturbations) and global (outliers) data contamination.
My research is supported by the NSF CAREER Award, an NSF DMS Grant, an NSF-BSF CIF Grant, and the IBM University Award.