Karl Pertsch

Karl Pertsch is a member of the technical staff at Physical Intelligence, where he works on training robot foundation models. He was previously a postdoctoral scholar at UC Berkeley and Stanford University, co-advised by Sergey Levine and Chelsea Finn, and completed his PhD at the University of Southern California under Joseph Lim. He is best known for his work on building generalist robot policies, including contributions to the Open X-Embodiment dataset, DROID, OpenVLA, and the FAST action tokenization method for vision-language-action models.