About the job
We're looking for a Simulation Engineer to build and own the simulation pipeline that powers AI training for our robots. You will create realistic 3D environments, place and configure our robot platforms inside them, and deliver high-quality synthetic data and evaluation scenarios that downstream learning teams can rely on.
Your work sits on the critical path between product hardware, autonomy, and model training: the environments you build need to be physically plausible, repeatable, and scalable enough to train policies that transfer to the real world.
What you'll do
- Design, implement, and maintain an end-to-end simulation pipeline for robot AI training (environment authoring → asset ingestion → robot placement → physics/sensors → data export)
- Build and iterate realistic 3D scenes that reflect real deployment contexts (layouts, objects, materials, lighting, clutter, and variation)
- Import, configure, and validate robot models, end-effectors, sensors, and calibration in simulation so they match hardware behavior closely enough for learning
- Define standards for scene composition, asset versioning, randomization, and reproducibility across training runs
- Generate and curate synthetic datasets (RGB-D, segmentation, proprioception, contact, and related modalities) for policy and VLA training workflows
- Partner with RL, perception, and deployment engineers to close the sim-to-real gap: identify failure modes, tune domain randomization, and improve evaluation suites
- Instrument simulations for benchmarking, regression testing, and continuous integration as the robot fleet and models evolve
- Document pipelines, scene libraries, and operational runbooks so others can extend and run simulations reliably
What we're looking for
- Degree in Robotics, Computer Science, or related field (or equivalent practical experience)
- Strong experience building 3D simulation environments and tooling for robotics, ML, or high-fidelity interactive systems
- Hands-on skills with one or more major simulation stacks (e.g. NVIDIA Isaac Sim / Isaac Lab, MuJoCo, Gazebo, Unreal Engine, Unity, or similar)
- Proficiency in Python and comfort working in Linux-based dev environments
- Experience with 3D asset workflows: meshes, materials, textures, collision geometry, and scene graph management (USD and similar formats are a plus)
- Understanding of robot kinematics, sensors (cameras, depth, IMU), and how simulated observations map to real robot stacks
- Ability to own cross-functional problems from requirements through delivery, with clear metrics for quality and throughput
- High agency, strong judgment, and a bias toward building and validating
Nice to have
- Experience with domain randomization, procedural scene generation, or synthetic data at scale for learning-based robotics
- Familiarity with ROS / ROS 2 integration in simulation (topics, TF, calibration, replay)
- Exposure to photorealistic rendering, ray tracing, or sensor noise models used for sim-to-real transfer
- Experience placing and tuning manipulators, mobile bases, or human-scale robots in complex environments
- Background in physics simulation tuning (contacts, friction, articulation limits, timestep stability)
- Familiarity with teleoperation data replay, digital twins, or fleet-log-driven scene reconstruction
Who you are
- You want simulation to be a force multiplier for real robots, not a demo
- You care about fidelity, repeatability, and how synthetic data compounds into field performance
- You are ambitious, fast-moving, and comfortable collaborating across robotics, AI, and hardware
- You like hard problems at the intersection of 3D content, physics, and learning systems
- You want to help define how intelligent machines are trained before they touch the real world
What we are looking to build
A production-oriented simulation pipeline that can spin up realistic environments, place our robots and tooling correctly, and export training-ready data for manipulation, navigation, and VLA policies—with evaluation harnesses that catch regressions early and inform sim-to-real improvements.
