Max Robotics — Global Service Robot Directory + US Certification

About

Eric Jang is a robotics and machine learning researcher whose work has profoundly shaped modern approaches to vision-based robotic manipulation, reinforcement learning, and self-supervised representation learning. His research consistently tackles one of robotics' most enduring challenges: enabling robots to learn dexterous, generalizable skills directly from visual perception with minimal human supervision. Jang's most celebrated contribution, QT-Opt (2018, 575 citations), demonstrated that scalable deep reinforcement learning could be applied to real-world robotic grasping at unprecedented scale, becoming a landmark reference in the field. Equally influential, his Time-Contrastive Networks work (2018, 555 citations) pioneered self-supervised learning from unlabeled multi-viewpoint video, opening new pathways for robots to learn by observing humans. His research on sim-to-real transfer—including RetinaGAN and recurrent visual servoing—addresses the critical gap between simulation training and real-world deployment, while Grasp2Vec contributed elegant object-centric representation learning without human labeling. More recently, BC-Z (2022) explored zero-shot task generalization through broad-scale imitation learning, signaling his growing interest in generalizable robot intelligence. Across more than a dozen publications accumulating over 1,600 citations, Jang has established himself as a leading voice in bridging perception, learning, and physical robotics.