Dr. Suraj Yadav

Suraj is working with Hertie AI Fellow Georgia Koppe. His research explores how geometric structures—manifolds of distributions, tangent spaces, curvature, and topology—govern learning dynamics in neural networks, diffusion models, and reinforcement learning. He is particularly interested in understanding synthetic data generation, world models in POMDPs, and topological constraints in nonlinear dynamical systems. His goal is to build mathematically principled foundations for modern AI by uncovering the geometry hidden inside data, optimization, and representation learning.