Research Intern Dexterous Manipulation & Embodied Vision (MS/PhD, 6-12 months)

Join our research team to explore learning-based control and perception systems for dexterous humanoid manipulation. You'll contribute to training high-DOF control policies, building vision-tactile fusion pipelines, and scaling data-driven methods to tackle real-world manipulation tasks.

Requirements: Currently enrolled in an MS or PhD program in Robotics, Computer Science, or related field

Strong foundation in reinforcement learning, imitation learning, or policy optimization

Experience with robot simulators (e.g., MuJoCo, Isaac Gym/Lab, Brax)

Proficiency in Python and at least one deep learning framework (PyTorch, JAX)

Experience with vision-based policy learning or contact-rich manipulation tasks

Familiarity with multi-modal sensor fusion (vision + proprioception + tactile)

(+) Experience with diffusion policies, VLPs (e.g., RT-1/2, VIMA), or behavior cloning from video

• (+) Hands-on experience with robotic hardware or sim2real adaptation methods

• Paid internship with competitive compensation

• Work on cutting-edge problems in dexterous manipulation and vision

• Mentorship from researchers and engineers working at the frontier of embodied intelligence

• Access to real robot hardware, simulation clusters, and large-scale datasets

• Opportunity to publish or contribute to high-impact research alongside product-driven development

• Design and evaluate policies for high-DOF dexterous hands in simulation and real-world settings

• Integrate multimodal sensor streams (RGB, depth, tactile, joint states) into control pipelines

• Experiment with vision-based pretraining, diffusion control models, or large-scale imitation learning

• Contribute to data collection tools and replay infrastructure for learning from human demonstrations

• Work closely with hardware, AI, and perception engineers to close the loop from sensing to control