SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Dyna Robotics is seeking a Research Engineer/Scientist to own the simulation pipeline powering their embodied AI foundation model for general-purpose wheeled mobile manipulators. You will build the simulated worlds that enable loco-manipulation (joint base and arm control) policies to generalize across real-world facility-scale environments.
Key responsibilities include: (1) Real-to-Sim Reconstruction—develop pipelines to reconstruct full facility-scale environments (rooms, aisles) into photorealistic simulation using 3D reconstruction and rendering; (2) Facility-Scale Scene Generation—procedurally generate navigable environments with varied layouts, obstacles, and object placements to stress-test loco-manipulation policies; (3) Sim-for-Data-Gen—generate synthetic training data for policies that jointly reason about base positioning and arm manipulation, incorporating real-world failure modes; (4) Loco-Manipulation Evaluation—build simulation benchmarks testing joint base+arm policies (VLA, imitation learning, diffusion models) holistically; (5) Rendering for ML—prioritize photorealistic rendering fidelity specifically to close the sim-to-real gap for vision-based control; (6) Cross-Team Collaboration—partner with AI Research and Data Ops teams to identify where simulated data and evaluation most accelerate policy development.
You should bring an MS/PhD in CS, Robotics, or Graphics (or equivalent hands-on depth), with demonstrated experience building simulation and rendering systems. Required: hands-on work with simulation stacks (MuJoCo, Isaac Sim, SAPIEN, Omniverse, Blender); procedural scene/asset generation and domain randomization at scale; experience training/evaluating manipulation or loco-manipulation policies; strong Python, PyTorch, or JAX skills. Highly valued: experience with mobile bases (wheeled, tracked, holonomic), real-to-sim-to-real pipelines, 3D reconstruction (NeRF/Gaussian splatting), world models, GPU-scale physics simulation, or prior work on wheeled mobile manipulators (e.g., Boston Dynamics Stretch). This is a founding-type ownership role requiring senior-level judgment to define research agenda and identify where simulation adds the most leverage, not a narrow IC slot on an existing pipeline. Publications at CoRL, RSS, ICRA, NeurIPS, CVPR, or SIGGRAPH, or experience leading small research teams, are bonuses.
