Physics-guided ML for computational imaging and 3D reconstruction. Develops provable algorithms, neural fields, and diffusion priors to solve inverse problems across medical imaging, scene understanding, and protein structure.