As Luma's founding Robotics Engineer, you'll bring up commercial robot platforms - humanoids, arms, mobile bases - wire up the sensors and data pipelines our world models need, and run the experiments that tell us whether the model actually works in the real world. We don't build robots here; we test models on them.
You'll be the bridge between research and the physical world, keeping a fleet of off-the-shelf robots running and debugging failures across hardware, drivers, and integration code. It fits a strong systems-integration engineer who's stood up complex robot setups and kept them alive. If you want to design robots rather than use them as a means to an end, this isn't that.
What You'll Own- Bring up and operate off-the-shelf robot platforms (humanoids, manipulators, mobile bases) for model evaluation and data collection.
- Integrate cameras, depth sensors, IMUs, and other sensors for world-model rollouts and policy evaluation.
- Own the data pipeline from robot to training: logging, synchronization, calibration, and replay.
- Build the test rigs and benchmarks that show whether the model is improving over time.
- Keep the fleet running, debugging failures across hardware, drivers, and integration code.
- Translate "what the model needs" into "what the robot does."
First 90 DaysOne way the first 90 could unfold.
- Days 1-30 - Immerse & Diagnose: Bring up the first robot platform and sensor stack, and map what world-model evaluation needs.
- Days 30-60 - Ship & Validate: Stand up a working robot-to-training data pipeline and a first benchmark.
- Days 60-90 - Scale & Systemize: Expand the fleet and make evaluation repeatable and reliable.
What You Bring- A strong systems-integration background: robots, drivers, sensors, real-time control.
- Hands-on with ROS/ROS2 (or equivalent), Python, and Linux at the systems level.
- Comfort debugging across the stack, from kernel-level driver issues to ML pipeline bugs.
- A track record standing up complex robot setups and keeping them running.
Nice to Have- Experience with humanoids or manipulators from major vendors (Unitree, Figure, 1X, Agility, Franka).
- Prior work in a research lab where the robot was a means to an end, not the end itself.
