Spring 2027 Robot Learning R&D Co-op

About the role

Spring 2027 Robot Learning R&D Co-op As part of the Research & Advanced Development group, you will help develop and integrate new approaches to solving a variety of challenging manipulation problems for real-world robotics systems, allowing them to understand and interact with their environment in unprecedented ways. This position offers a unique opportunity to work at the cutting edge of robotics applied to real-world challenges. Your work will span the full range of this robotics stack, from robot drivers to the multi-robot orchestration system running in Berkshire Grey’s in-house robot proving ground. This role will be onsite at the Bedford, MA office.  This position will run from approximately January to June 2027 (dates flexible). Functions: • Build software infrastructure to integrate, deploy, and evaluate learning-based manipulation capabilities • Contribute to the training and adaptation of existing machine learning architectures for robot learning and visual scene understanding • Contribute to the development and continuous improvement of Berkshire Grey’s robot learning data collection application • Build tools to process, validate, visualize, and evaluate data from on-robot policy execution, including system logs, sensor data, annotations, and failure cases • Integrate novel grippers and other hardware into existing perception, planning, and robot-control systems • Leverage production data and our internal development systems to implement and evaluate your solutions Qualifications • Currently pursuing an advanced degree (MS or PhD) in Robotics, Machine Learning, Computer Science, or a closely related field • Strong software development expertise in Python; C++ is a plus • Experience with modern machine learning methods such as imitation learning, reinforcement learning, generative models, vision transformers, or related methods • Experience applying ML models to systems that interact with the physical world • Experience developing robotics software using Linux (Ubuntu) and ROS COOP-2621