Max Robotics — Global Service Robot Directory + US Certification

About

Serkan Cabi is a machine learning researcher specializing in deep reinforcement learning, robotic manipulation, and imitation learning, with a particular focus on bridging the gap between learning from demonstrations and autonomous agent training. His most influential work, "Reinforcement and Imitation Learning for Diverse Visuomotor Skills" (2018), garnered over 330 citations across versions and pioneered a model-free approach that combines modest demonstration data with reinforcement learning to train end-to-end visuomotor policies directly from RGB inputs — a significant step toward practical robotic learning. Cabi further advanced the field with his data-driven robotics framework, which scales robotic manipulation tasks using learned reward functions and batch reinforcement learning, accumulating over 100 citations. His research into reward learning — including reward sketching, semi-supervised reward learning, and vision-language models as success detectors — reflects a sustained effort to make reward specification more scalable and generalizable. Additional contributions, such as the Intentional Unintentional Agent for multi-task continuous control and work on adversarial imitation learning robustness, demonstrate the breadth of his expertise. Cabi's body of work has meaningfully shaped how researchers approach sample-efficient, real-world robot learning.