Max Robotics — Global Service Robot Directory + US Certification

About

David Barker is a robotics researcher specializing in data-driven approaches to robot learning, with a particular focus on reinforcement learning, visual perception, and robot manipulation. His most influential work centers on developing scalable frameworks that enable robots to learn from large datasets of recorded experience, eliminating the need to engineer tasks from scratch. His paper "Scaling Data-Driven Robotics with Reward Sketching and Batch Reinforcement Learning," which has accumulated over 100 citations across versions, demonstrates how learned reward functions can guide robots through multiple object manipulation tasks on real hardware — a significant step toward practical, generalizable robot learning systems. Barker has also made meaningful contributions to visual representation learning for robotics. His work on S3K (Self-Supervised Semantic Keypoints) introduces a multi-view consistency approach to learning structured visual representations without manual annotation, addressing a fundamental bottleneck in robotic perception. Additionally, his research on deep reinforcement learning for variable-socket insertion tasks highlights his interest in solving industrially relevant manipulation challenges. Collectively, his work bridges the gap between large-scale machine learning and real-world robotic deployment, making him a noteworthy contributor to the growing field of data-driven robotics.