Max Robotics — Global Service Robot Directory + US Certification

About

Dibya Ghosh is a leading researcher at the intersection of robotics, reinforcement learning, and foundation models. Her work focuses on building generalist robot policies that can learn from diverse, large-scale datasets and generalize to real-world tasks without task-specific training. Ghosh is best known for her central role in developing **Octo**, an open-source generalist robot policy pretrained on a massive corpus of robot demonstration data. This model, which has already garnered over 70 citations, demonstrates that a single policy can be finetuned with minimal in-domain data to perform a wide variety of manipulation tasks, representing a paradigm shift away from training robot policies from scratch. She has also pioneered methods for **robotic offline reinforcement learning from Internet videos**, showing how value-function learning can leverage passive video data to improve robot control, and has contributed foundational work in **distributionally adaptive meta-RL**, enabling policies to generalize to unseen task distributions. Most recently, Ghosh co-authored **π₀.₅**, a vision-language-action model designed for open-world generalization, pushing the boundaries of how far end-to-end robot learning can operate outside the lab. Her contributions are shaping the future of scalable, general-purpose robotics.