The NVIDIA Isaac Loco-Manipulation team is seeking exceptional machine learning Interns to join our world-class robotics initiatives. As an intern, you’ll work alongside industry-leading experts, gaining hands-on experience and contributing to the future of humanoid loco-manipulation. We’re looking for strategic, ambitious, and creative individuals who are passionate about pushing the boundaries of robotics.
What You’ll Be Doing:
Collaborate with researchers and engineers to define and execute projects focused on humanoid loco-manipulation and mobile manipulation.
Contribute to GR00T and Cosmos foundation models.
Support the development of reference workflows in Isaac Lab and Newton.
Advance technologies for robot learning and synthetic data generation using human video datasets.
Design, implement, and deploy novel algorithms for humanoid robot locomotion and manipulation in both simulated and real-world environments.
Integrate your work with NVIDIA’s advanced robotics platforms.
Transfer your innovations into products, with deliverables including prototypes, patents, and/or publications in top conferences and journals.
What we need to see:
Currently enrolled in a PhD or Master’s program in Computer Science, Electrical Engineering, Robotics, or a related field, and available for the duration of the internship.
Strong programming skills in Python and C++; familiarity with deep learning frameworks (PyTorch, JAX, TensorFlow) and physics simulation tools (Isaac Sim/Lab, MuJoCo).
Demonstrated research or internship experience, with publications in top conferences.
Excellent communication and collaboration skills.
Experience with large-scale model training on GPU clusters is a plus.
Ways to stand out from the crowd:
Foundation models for robotics and 3D perception.
Learning from human video demonstrations; Human-object reconstruction.
Humanoid loco-manipulation: whole-body control, dexterous and bimanual manipulation, locomotion.
Robotics simulation, sim-to-real and real-to-sim transfer.
Robot learning and reasoning, including imitation and reinforcement learning:
Vision-language-action (VLA) models.
Synthetic data generation for robotics research.