Max Robotics — Global Service Robot Directory + US Certification

About

Karl Pertsch is a robotics and machine learning researcher whose work sits at the intersection of large-scale robot learning, foundation models, and generalist robot control. He has made significant contributions to some of the most influential robotics projects of the past few years, including the landmark RT-1 and RT-2 papers (512 and 267 citations respectively), which demonstrated how transformer-based architectures and vision-language models trained on internet-scale data can be leveraged to dramatically improve robotic generalization and real-world performance. Pertsch has been a key contributor to the development of open-source generalist robot policies, including Octo and OpenVLA, democratizing access to powerful robotic foundation models. His work on the DROID large-scale manipulation dataset and the π₀ vision-language-action flow model reflects a sustained commitment to building robust data infrastructure and scalable control frameworks. Earlier work on skill priors for reinforcement learning highlights his foundational interest in knowledge transfer and sample-efficient learning. More recent contributions, such as FAST action tokenization and language-correction methods for robots, demonstrate his range across both practical deployment and core algorithmic innovation. Across his career, Pertsch has helped define how modern AI techniques are reshaping robot learning at scale.