Building the general-purpose robotic brain
Skild AI’s approach to a shared foundation model for robot perception, navigation, and manipulation.
Shagun Uppal
I’m building toward scientific superintelligence at Warp.
“The test of all knowledge is experiment.”
— Richard Feynman
Before Warp, I worked on general-purpose embodied intelligence at Skild AI. I did my MS in Robotics at Carnegie Mellon with Deepak Pathak, studying how active vision enables reactive behavior: robots learning where to look and using visual feedback to coordinate navigation and manipulation in real time.
In a past life
My research journey started at IIIT Delhi, where I studied Computer Science and worked with Saket Anand and Pavan Turaga. Along the way, I also spent time at USC as a visiting researcher in Joseph Lim’s CLVR Lab.
Selected research
All publications ↗
The case for an omni-bodied robot brain
Learning to adapt across different robot bodies, hardware failures, and unfamiliar physical conditions.
SPIN: Simultaneous Perception, Interaction and Navigation
SPIN brings perception, interaction, and navigation into one coordinated robot system.
Dextrous Functional Grasping
Learning to grasp everyday objects in ways that let a robot actually use them.
Robots, science, or a book I should read?
Say hello