NVIDIA is looking to solve the problem of building foundation models for robotics, specifically in the area of superhuman dexterity, by discovering, refining, and building the training infrastructure to unlock the potential of dexterous robots around the world.
Requirements
- Strong Pytorch and training experience, especially with large-scale systems.
- Experience working with robotic hardware, in particular manipulators and multi-fingered hands.
- Experience with Isaac Lab/Sim/Gym RL training and experience with Warp.
- Deep knowledge of modern perceptual backbones and internet scale robotics foundation model training.
- Experience with tactile sensors, physical and simulated, and multimodal sensorimotor processing.
- Strong understanding of mathematical concepts related to robotics and AI such as SE(3) transforms, kinematics, dynamics, control, numerical differential equations, nonlinear geometry, high-dimensional geometry, kernel methods, diffusion models.
Responsibilities
- Scaling RL training in sim for real-world deployment, and developing new scalable RL training paradigms for training skills foundation models.
- Studying high-dimensional sensor processing and multi-modal integration (stereo RGB, depth, tactile, proprioception).
- Working on bleeding edge sim-to-real dexterity training pipelines deployed to humanoids (upper torso and full body).
- Rigorous scientific analysis of tradeoffs in training methodologies.
- Coordinating closely with the simulation team on scalable simulation infrastructure and sensor simulations for training.
- Collaborate with other research teams throughout the company and transfer technology to product teams.
Other
- MS or PhD degree in a robotics related field (or equivalent experience).
- 4+ years of work related experience in AI and learning approaches to robotics, especially RL and/or modern imitation learning.
- Creativity in solving problems and ambition to work steadily toward long-term visions.
- Base salary range is 184,000 USD - 299,000 USD.
- Eligible for equity and benefits.
