Robotics Core pushes the algorithmic boundary in navigation, localization and planning. Frontier turns robot state into interfaces people actually trust. You sit between them and turn what they build into a robot that shows up for its shift, does its job, and doesn't page anyone at 3am. Not a hardware role: no motor drivers, no sourcing LiDAR units. Everything above that is yours, from the driver layer to a fleet of dozens moving through our Master Control System (VDA5050-compliant).
You're not maintaining one module. You're accountable for how our robots perform in the field.
What you'll do
- Sit between Robotics Core and Frontier: Their output becomes something operable, field reality becomes requirements they can act on. What's worth building and what's worth working around is your call.
- Own the driver-to-behavior stack: Hardware abstraction, ROS 2 DISTRO Jazzy driver layer, state machines. Where the line falls between a hardware quirk you abstract away and a behavior bug you fix properly is your decision, and every later feature inherits it.
- Design and guard the safety-relevant logic on the robot: Every rule trades throughput against certainty, and you own the trade-off, not just the code that implements it.
- Own robot-to-fleet communication: VDA5050 up to the Master Control System, plus the coordination and traffic-management logic that has to hold up on a real warehouse floor. The algorithms often come from Robotics Core, but when integration turns up gaps you dig into them instead of just wiring them up.
- Run the Master Control System as a production service: Cloud Deployed using Docker. Observability using OpenTelemetry, Orchestration open. You define what counts as a healthy fleet, and what triggers an alert versus a shrug.
- Own deployment and SRE for robot and fleet software: Every rollout strategy, rollback plan and monitoring threshold is a bet on how much operator trust you'll risk for speed.
- Debug production incidents on live customer fleets: Whether you call something a root cause or a symptom decides whether the same failure shows up again next week.