In this talk, Juliàn Tachella will present his recent work on computational imaging theory and applications. On the applications side, he will introduce a new computational framework for recovering 3D point clouds in real-time from noisy and incomplete single-photon lidar measurements, and a novel non-line-of-sight imaging technique, which relies on the co-design of sensing and computation. On the theoretical side, he will present an analysis of the implicit bias of convolutional neural networks towards clean images using recent advances on the large system limit interpretation of neural networks. His analysis identifies strong links between CNN architectures and well-known signal processing techniques such as non-local means.