Robotics Simulation & Synthetic Data Intern
Core
Build end-to-end synthetic data generation pipelines for robot foundation models using NVIDIA Isaac Sim and Cosmos Transfer to create photorealistic training datasets from human demonstrations.
Role type
Robotics Simulation & Synthetic Data Intern
Builds
Large-scale synthetic trajectory and image datasets for VLA (Vision-Language-Action) model training
Domain
Robotics / Physical AI / Embodied AI
Deliverable
production ML models
Required skills
Python, Linux/Docker, GPU workloads (CUDA), physics simulation (NVIDIA Isaac Sim/Lab, MuJoCo, PyBullet, Gazebo), robot learning fundamentals (imitation learning, behavior cloning, RL), USD/URDF, ROS/ROS2
Preferred skills
NVIDIA Omniverse, Replicator, VLA architectures (RT-2, Octo, π0, GR00T N1), domain randomization, neural rendering (NeRF, Gaussian Splatting)
Technologies
NVIDIA Isaac Sim, Isaac Lab, GR00T-Mimic, Cosmos Transfer, USD, URDF, ROS/ROS2, Open X-Embodiment, LeRobot, ACT, Diffusion Policy
Responsibilities
Configure simulation environments with scene composition, physics parameters, and domain randomization; apply world-foundation-model augmentation for photorealistic rendering; design and run data quality experiments; curate and version datasets; benchmark synthetic vs. real data performance; document pipelines for team reproduction
Seniority
Intern
