Max Robotics — Global Service Robot Directory + US Certification

About

Shubham Tulsiani is a robotics and computer vision researcher whose work sits at the intersection of robot learning, visual perception, and generalizable manipulation. His research is driven by a central challenge: enabling robots to interact meaningfully with the physical world without relying on exhaustive, expensive data collection or hand-engineered systems. Tulsiani has made significant contributions toward data-efficient robot learning, pioneering approaches that leverage passive human videos, semantic augmentations, and pre-trained visual representations to train manipulation policies that generalize across novel objects and settings. His work on RoboAgent (70 citations) demonstrated that combining semantic augmentations with action chunking can yield surprisingly capable and generalizable manipulation systems despite limited robotics data. His research on zero-shot manipulation — translating human interaction plans and point tracks from internet videos into robot policies — has opened promising avenues for reducing the robot-human data gap without costly teleoperation. Earlier contributions explored object-centric forward modeling and the discovery of recomposable motor primitives from diverse demonstrations. Across his body of work, Tulsiani consistently pursues a compelling vision: robots that learn broadly and cheaply, much like humans do, by observing the world around them.