PhD Research Intern, Learning Embodied Skills from Human Data - 2027

NVIDIA is seeking outstanding Research Interns to join the Data-Driven AI for Robotics (DAIR) group. The focus is on learning embodied skills from large-scale human data. Our objective is to develop AI systems that capture, understand, and reproduce complex human motion and interaction skills across physical and digital embodiments, including humanoid robots and animated characters.

Our research spans the full stack: reconstructing human motion and human-object interactions from video; generating diverse, controllable character behaviors; transferring motion across embodiments; and training physically grounded controllers for humanoid robots and interactive virtual characters.

You will collaborate with a passionate and supportive research team that consistently produces influential work published at leading computer vision, machine learning, graphics, and robotics conferences. You will also have the opportunity to collaborate with world-class research and product teams across NVIDIA, following our strong “one-team” culture.

What you'll be doing:

• Innovate and implement novel AI algorithms that transform large-scale human data into controllable motion and interaction skills across physical and digital embodiments.
• Develop robust, scalable training and inference pipelines for motion reconstruction, generation, retargeting, and character and robot control.
• Build methods that transfer human skills to humanoid robots, including whole-body loco-manipulation and dexterous manipulation.
• Maintain a close, collaborative relationship with your mentor(s).
• Publish your research findings at leading computer vision, machine learning, graphics, and robotics conferences.
• Partner with product teams to enable effective technology transfer of your work.

Research Topics Include:

• Human motion and human-object interaction reconstruction, synthesis, and generation.
• Learning character and robot skills from video, motion-capture, and teleoperation data.
• Cross-embodiment motion generation, retargeting, and tracking.
• Whole-body humanoid control, loco-manipulation, and dexterous manipulation.
• Reinforcement learning and imitation learning.
• Differentiable physics simulation and physically grounded motion generation.
• World action models, vision-language-action models, video and motion foundation models, and LLM-based agents for data generation and embodied AI.

What we need to see:

• Pursuing a PhD degree in Computer Science, Computer Engineering, Electrical Engineering, Robotics, or a related field.
• Highly efficient and creative use of coding agents to accelerate research prototyping, experimentation, and development.
• Outstanding engineering skills in rapid prototyping and developing model-training and simulation frameworks (e.g., PyTorch, Isaac Lab, MuJoCo).
• Excellent skills in working with large-scale machine learning/AI systems and compute infrastructure.
• A promising research track record with at least one publication at leading computer vision, computer graphics or robotics conference (e.g. CVPR, ICCV, SIGGRAPH, CoRL, etc.).
• Experience in human motion modeling, human-object interaction, character animation, robot learning, reinforcement or imitation learning, cross-embodiment motion tracking, generative modeling, video understanding, or differentiable physics simulation is preferred.
   

NVIDIA is widely considered to be one of the technology world’s most desirable employers. We have some of the most forward-thinking and productive people in the world. Please join us and be at the forefront of developing the next generation of embodied AI systems, spanning humanoid robots and digital characters.

You will also be eligible for Intern benefits. ​

This posting is for an existing vacancy. 

NVIDIA uses AI tools in its recruiting processes.