Quan Vuong

Quan Vuong is a co-founder of Physical Intelligence, developing robotic foundation models that transfer across machines, tasks, and environments. His research helped establish how robots can share learned capabilities instead of requiring separate, narrowly programmed control systems.

Vuong pursued doctoral research in deep reinforcement learning, computer vision, and robotics at the University of California, San Diego, with research internships at Microsoft Research, Google Research, and Google Brain. At Google DeepMind, he contributed to RT-1 and RT-2, which connected large-scale robotic control with knowledge acquired from vision-language training.

In 2023, he and Pannag Sanketi introduced Open X-Embodiment and RT-X, a collaboration involving 33 academic laboratories and 22 robot types. Its central finding—that experience pooled across different machines can improve individual robots’ performance—gave cross-embodiment learning a practical foundation.

At Physical Intelligence, Vuong helped develop π0, a generalist robot policy capable of dexterous manipulation, and contributed to policy research and infrastructure for π0.5 and π0.7. These efforts extend robotic learning into unfamiliar homes, longer household tasks, and more adaptable foundation models.

• Vision-language-action models: Integrate visual understanding, language, and robot control so broadly pretrained systems can execute physical tasks.
• Open-world robotic generalization: Train across diverse environments and tasks to help robots handle unfamiliar homes, objects, and multistep instructions.
• Hardware-independent robot intelligence: Make learned policies portable across different machines; one coffee-making demonstration used a partner’s robot that Vuong’s team had never physically accessed.
• Cloud-hosted robotic models: Provide shared intelligence that specialized robotics companies can apply to particular customers and operational settings, an approach Vuong outlined in a Y Combinator conversation.