DYNA Robotics
Research Scientist, Simulation
Redwood City, CA · On-site · Full-time · Posted Aug 4, 2026
- Experience
- No fixed years-of-experience bar; demonstrated depth building simulation/rendering systems is emphasized.
- Education
- MS or PhD in CS/Robotics/Graphics, or equivalent hands-on experience.
About the job
Build end-to-end simulation systems, environments, training data, and benchmarks for wheeled mobile-manipulation policies, with emphasis on facility-scale reconstruction and photorealistic rendering.
What you'll do
- Reconstruct real facility-scale environments and build procedural scenes for simulation.
- Generate synthetic training data and evaluation benchmarks for policies jointly controlling base motion and arm manipulation.
- Improve rendering fidelity and collaborate with AI Research and Data Ops to guide simulation and real-world data efforts.
What they ask for
- Hands-on experience with robotics or graphics simulation stacks such as MuJoCo, Isaac Sim/Isaac Lab, SAPIEN, Omniverse, or Blender.
- Experience with procedural scene or asset generation, domain randomization, or scalable photorealistic rendering pipelines.
- Familiarity with training or evaluating manipulation or loco-manipulation policies and how simulated data supports them.
- Strong Python skills; comfort with PyTorch or JAX for model training or evaluation.
Nice to have
- Experience simulating mobile bases or prior research or engineering in loco-manipulation or whole-body control.
- Real-to-sim-to-real pipelines, world-model or video-prediction research, GPU-scale physics simulation, wheeled mobile manipulators, relevant publications, or research-team leadership/mentoring.
Tools
MuJoCo, Isaac Sim/Isaac Lab, SAPIEN, Omniverse, Blender, Python, PyTorch, JAX
AI focus
Develop simulation-generated data and benchmarks to train and evaluate vision-based loco-manipulation policies, including VLA, imitation learning, diffusion, or RL approaches.
Model-written summary · the original posting is authoritative