Research centers on computational models of attention and perception via information-theoretic and deep learning methods, spanning saliency prediction, semantic segmentation, and spatiotemporal representation learning; recent work quantifies static-versus-dynamic biases and enables pixel-level editing in 3D neural scenes.