Role Description

We are looking to recruit an exceptional Robot Autonomy Engineer to build the decision-making stack that turns a goal into coordinated, reliable robot behavior in real industrial applications — what the robot should do next, in what order, and how a fleet of them shares a workspace without getting in each other's way.

In this role you will:

• Own the autonomy stack above the controller — task planning, behavior planning, path planning and trajectory planning — from the moment work arrives to the trajectories handed off to motion control.


• Design the behavior architectures that structure long-horizon manipulation and navigation tasks, and that degrade into retry, recovery and operator handoff rather than into a stall.


• Bring principled task planning to industrial workflows: goal and precedence reasoning, task allocation, and planning under uncertainty.


• Plan and coordinate motion for multiple robots sharing an industrial facility — separation, reservation, deconfliction and deadlock-free repositioning — so that adding a robot adds throughput.


• Integrate LLM and VLM reasoning into planning for task decomposition, subtask grounding and language-conditioned goals, together with the verification and fallbacks that make a model's output safe to execute on real hardware.


• Define the contract between learned policies and classical planning: what the model may decide, what the planner must guarantee, and how the two hand off mid-task.


• Interface with perception, intelligence, controls, simulation and platform software in designing functional architectures that hold up under real-world operation.


• Hold the whole stack to measurable field performance — cycle time, success rate, intervention rate — through simulation, replay of recorded robot logs, and testing on real robots.

Qualifications

Must-have:

• MS or PhD in robotics, engineering, mathematics, computer science or a related discipline.


• Real-world experience in classical motion planning for one or more robots — search-based, sampling-based or optimization-based (A*, RRT/PRM, trajectory optimization, model predictive control) — carried onto hardware rather than left in simulation.


• Real-world experience in behavior planning: finite state machines, behavior trees or comparable behavior architectures for long-horizon tasks, including failure detection and recovery.


• Familiarity with task planning in the classical AI planning sense (STRIPS, PDDL, HTN) or decision-theoretic planning (MDP, POMDP), and the judgment to know when that machinery earns its complexity against a simpler reactive design.


• Proficiency in Python and C++ programming, using up-to-date software development practices and tooling.


• Self-starter attitude with strong ability to identify problems, prioritize them, then plan and execute working solutions.


• Enthusiasm for working in a fast paced startup environment and eagerness to support the team on a variety of topics.

Nice-to-have:

• Practical experience fine-tuning and integrating LLMs or VLMs for task planning, including grounding model output in executable, verifiable plans.


• Multi-robot coordination at fleet scale: task allocation and assignment, traffic management, deconfliction, multi-agent path finding.


• Experience with mobile manipulation — coordinating a mobile base and one or more arms toward a single task.


• Familiarity with ROS 2, and with fleet interface standards such as VDA5050.


• Familiarity with planning and kinematics libraries such as Drake, OMPL or MoveIt.


• Experience evaluating planners in simulation and against replayed field logs, and the regression testing that keeps a planner honest as it changes.


• A track record of carrying autonomy from working demo to sustained field operation.


• Familiarity with functional safety (FuSa) concepts.