Works on diffusion-based generative modeling, multi-modal vision, and physics-informed learning for dynamical systems. Applies transformers and RL to perception and control for manufacturing quality, scientific simulation surrogates, and robotics.