Max Robotics — Global Service Robot Directory + US Certification

About

Albert Zhan is a robotics and machine learning researcher whose work sits at the intersection of robot learning, representation learning, and data-driven policy development. He is perhaps best known as a contributor to **DROID**, a landmark large-scale in-the-wild robot manipulation dataset (2024, 108 citations), which addresses one of the field's most pressing challenges: collecting diverse, high-quality robot manipulation data across varied real-world environments. This work represents a significant milestone in scaling robot learning infrastructure. Zhan has also made meaningful contributions to improving the sample efficiency of real-robot reinforcement learning. His work on contrastive pre-training and data augmentation demonstrated how unsupervised representation learning techniques — previously successful in simulation — could be adapted to the demanding constraints of physical robot systems. Beyond single-skill learning, he has explored hierarchical imitation learning through skill transition models, enabling agents to tackle long-horizon tasks and generalize to novel scenarios. His work on adversarial policy ensembles further showcases his range, addressing the underexplored problem of policy privacy in deployed systems. Across these contributions, Zhan's research consistently pushes toward robots that learn more efficiently, generalize more broadly, and operate more reliably in the real world.