Max Robotics — Global Service Robot Directory + US Certification

About

David Khosid is a roboticist whose research sits at the intersection of computer vision and reinforcement learning, with a focus on enabling robots to learn complex manipulation skills directly from visual data. His most impactful work, "Self-Supervised Sim-to-Real Adaptation for Visual Robotic Manipulation," has accumulated 56 citations and addresses a critical bottleneck in robotics: the difficulty of obtaining reward signals from real-world visual data. Khosid’s key contribution is a self-supervised framework that leverages unlabeled data to bridge the "sim-to-real" gap, allowing policies trained in simulation to adapt to physical robots without costly human annotation. This approach dramatically accelerates robotic learning by eliminating the need for hand-crafted rewards. In a follow-up study, "Beyond Pick-and-Place: Tackling Robotic Stacking of Diverse Shapes" (16 citations), Khosid tackled a more challenging manipulation task, designing a diverse set of geometrically complex objects that require sophisticated, non-grasping strategies. His work demonstrates that vision-based reinforcement learning can handle real-world variability, pushing the field beyond simple pick-and-place operations. Khosid’s contributions are paving the way for more autonomous, adaptable robots in manufacturing and service settings.