Reinforcement Learning Intern

Job description

Job responsibilities

• Collect tele-operated simulation and/or real data from the surgical robot while performing a realistic exercise (e.g., Fundamentals of Laparoscopic Surgery pick-and-place).
• Train off-the-shelf reinforcement learning algorithms (e.g., policy gradient methods, actor-critic methods, etc.) on the trajectory data and benchmark their performance using established metrics.
• Adapt reinforcement learning algorithms to the surgical domain and exercise of interest and identify techniques for improving performance relative to off-the-shelf RL models.
• Recommend a suggested algorithm / experimental setup moving forward (beyond the internship) based on experimental results, and present findings to the broader team.
• Deploy chosen RL algorithm either in simulation or on the dV5 to validate its performance in real-time (vs. purely offline evaluation).

Job requirements

• University Enrollment: Must be currently enrolled in and returning to an accredited degree-seeking academic program after the internship.
• Internship Work Period: Must be available to work full-time (approximately 40 hours per week) during a 10-12 week period starting May or June. Specific start dates are shared during the recruiting process.
• Deep experience with a deep learning framework (e.g., PyTorch or JAX).
• Strong familiarity with reinforcement learning algorithms (e.g., policy gradient methods, actor-critic methods).
• Willingness to learn about robotics and the application of RL to surgical robotics.
• Excel in communicating, both in written and verbal form, about motivation for running experiments, experimental findings, and the implications of the work.
• Experience with video-based and multimodal data analysis and processing.