Max Robotics — Global Service Robot Directory + US Certification

About

Kevin Zakka is a robotics researcher whose work spans robot learning, dexterous manipulation, and control, with a focus on building systems that generalize beyond narrow, task-specific constraints. He is perhaps best known for **Form2Fit** (2020, 109 citations), a landmark contribution to robotic assembly that replaces reliance on 3D CAD models and pose estimation with learned shape priors, enabling robots to generalize assembly policies to previously unseen objects. This work demonstrated that disassembly could serve as a scalable supervisory signal — an elegant and practical insight for real-world robotics. Zakka has also made significant contributions to model-based control through **MuJoCo MPC** (2022), an open-source framework enabling real-time predictive behavior synthesis, and to **MuJoCo Playground** (2025), which streamlines sim-to-real transfer for robot learning. His work on **XIRL** (2021) advanced cross-embodiment imitation learning from human video, while **RoboPianist** (2023) pushed the frontier of dexterous reinforcement learning using piano playing as a benchmark. His involvement in **ALOHA 2** (2024) reflects a commitment to democratizing robot hardware. Across these projects, Zakka consistently bridges rigorous research with open, accessible tools for the broader robotics community.