Hi, my name is Sahit! I'm a final year Ph. D. student at MIT in the Learning and Intelligent Systems Group, advised by Leslie Pack Kaelbling and Tomás Lozano-Pérez.

During my Ph.D., I've spent time interning at the Robotics Lab at Autodesk and the Robotics Lab at DeepMind. Prior to coming to MIT, I did my undergrad at Georgia Tech, where I was advised by Byron Boots in the Robot Learning Lab. During undergrad, I also interned at the Human-Centered Robotics Lab at UW and the Search-Based Planning Lab at CMU.

Email  /  Resume  /  CV  /  Scholar  /  Github  /  Twitter
Goodreads  /  Letterboxd

I am interested in making robots reason under uncertainty, especially for tasks that involve making and breaking contact with the environment. My research focuses on applying ideas from motion planning, geometry, and machine learning to this end in the context of robot manipulation.

We present TiPToP, an extensible modular system that combines pretrained vision foundation models with an existing Task and Motion Planner to solve multi-step manipulation tasks directly from input RGB images and natural-language instructions.

We consider assembly tasks where the relative configuration parts is unknown due to sensing error. We formulate this problem as a search in Belief Space and introduce a Bi-Level Planner that reasons over both gripper trajectory and robot compliance.

We introduce a heuristic which encourages A* search to stay within a homotopy class extracted from human demonstration, which rapidly improves search runtime.

We show that sampling factor graphs can be used to recover diverse robot trajectories, which can be switched between in realtime for environments with moving obstacles.

Website Template from Jon Barron, but egregious design choices were made at Sahit's prerogative.