Mowze's commercial robotic lawn mowers are deployed in the field. However, a skilled operator on a manned mower still has a speed edge over our product, and we believe most of that gap is in the controller. You will own the robot's motion stack, covering state estimation, planning, and control. That includes building and refining a dynamics model of the mower from real telemetry, and using it to drive smoother and more precisely on slopes and uneven turf. Your goal is to exceed a human's efficiency. Hybrid, based in Gaithersburg, MD. On site 2-5 days a week, as needed.

Responsibilities Own the motion stack:

state estimation → planning → control → actuator commands. Choose the architecture and own the migration. Build, refine, and validate a dynamics model of the mower (mass properties, tire/turf interaction, slope effects, controller response) and fit it to real telemetry. Design model-based control (MPC or equivalent) that holds line on slopes and varying turf, slows for sharp corners, and maximizes ground speed subject to cut quality and safety limits.

Own coverage planning:

detect missed or partially cut strips and re-plan to cover them. Optimize path for throughput. Stand up simulation-in-the-loop and hardware-in-the-loop test infrastructure. Define what telemetry the mower must log to close the loop on model fidelity, work with other engineers to implement logging. Qualifications 3+ years shipping control systems on physical vehicles or mobile robots — ground vehicles, agricultural/construction equipment, AMRs, off-road platforms. Not drones-only, not simulation-only. Demonstrated system identification on real hardware: has fit a model to logged data, found it was wrong, and fixed it. Hands-on with model-based control in production (MPC, LQR/iLQR, or equivalent) — not just a course project. Strong state estimation (EKF/UKF, IMU + GNSS + wheel odometry fusion). Comfortable in C++ and Python; has worked inside a real-time embedded control loop, not only in a notebook. Has built a vehicle control stack from scratch or substantially rewritten one; understands what ArduPilot/PX4 do and do not model, so the replacement doesn't lose the parts that worked. Has built or substantially extended a vehicle simulator and can articulate where sim diverged from reality.

Pay:

$130,000.00 - $180,000.00 per year People with a criminal record are encouraged to apply Application Question(s): Do you have real world experience with a production fleet of robots / vehicles? (How many were in the fleet? Did the robots operate outdoors? Roughly how much did the robots weigh? What was and how long was your role?) How do you model a vehicle that is crabbing while driving sideways on a slope? While not sliding completely, there is some nonzero sideways slip. How do you fuse visual odometry, wheel odometry, IMU, and intermittent GPS for accurate pose estimates in GPS poor environments? About how far could the robot travel without GPS while still staying on the path? (State your assumptions about the system) Describe a time a model you identified from real data turned out to be wrong. How did you find out, and what did you change?

Work Location:

Hybrid remote in Gaithersburg, MD 20879