Shubham Tulsiani

Shubham Tulsiani is an Assistant Professor at Carnegie Mellon University's Robotics Institute, where he leads the Computer Vision group and the Physical Perception Lab. His research focuses on inferring physically and spatially grounded representations from perceptual inputs, with applications in 3D vision, robot manipulation, and neural scene reconstruction. He directs an active research group with multiple PhD and Master's students.

Research interests center on 3D scene understanding, robot learning, and generative modeling, with specific emphasis on: self-supervised perception, neural rendering, multi-view geometry, manipulation from visual inputs, and physics-based reasoning. The lab develops methods that leverage physical world constraints as supervisory signals.

Recent publications demonstrate strong focus on diffusion models for 3D tasks, sparse-view reconstruction, and robotic manipulation transfer. Key trends include neural inverse rendering, view synthesis from limited observations, and translating human interactions to robot actions.

Awards include:

• Best Student Paper Award at CVPR 2015

Advising includes supervision of 5 PhD students, 4 MS students, and undergraduates. Lab alumni hold positions at Google, Stanford, Meta, and Princeton. The Physical Perception Lab collaborates with FAIR Pittsburgh and the CMU Computer Vision group.