Uksang Yoo (유욱상)

I am a postdoctoral researcher with Prof. Ken Goldberg in the AUTOLab at UC Berkeley. I completed my Ph.D. at Carnegie Mellon University in the Robotics Institute (RI), where I was co-advised by Prof. Jean Oh and Prof. Jeffrey Ichnowski as a member of RI’s BIG lab and Momentum lab. I have also worked as a research scientist intern with the robot learning groups at Bosch AI (Summer 2025) and Meta (Fall 2025).

My research investigates how mechanics and physics-based models can enable generalizable learning for dexterous, contact-rich manipulation. In particular, I develop models of contact dynamics and deformation mechanics that provide strong inductive biases for robot learning. My work spans compliant and dexterous manipulation, learning from human demonstrations across embodiments through contact, tactile sensing, and reasoning about the continuous deformation of soft objects/robots during physical interaction.

This research draws on ideas from discrete differential geometry, machine learning, robotics, and soft-body mechanics, with the goal of deploying dexterous robotic systems reliably in the real world.

I received my B.Sc. in Mathematics (2020) and in Mechanical Engineering (2021) at the University of Texas at Austin, where I developed my interest in continuous deformations and continuum structure representation. My research is supported by the NSF Graduate Research Fellowship Program (GRFP).

Ph.D. Thesis · Carnegie Mellon University, 2026Mechanics-Informed Learning for Deformation-Rich ManipulationTL;DR: We already know how soft bodies deform (mechanics, contact modeling) but how can we encode them into general manipulation policy learning frameworks to learn to leverage deformation in our world with better sample efficiency and generalization?Read the dissertation →