Autonomy Engineer

QAIL AI focuses on building autonomously operated AI systems that function reliably in real-world environments. The team is engaged in deep applied research, training advanced algorithms on real-time data streams. QAIL AI aims to bridge cutting-edge artificial intelligence with practical deployment across physical systems and devices. We are building an autonomous ground vehicle for rough, cross-country terrain, and we need a founding engineer to own its autonomy, from the sensors to the path it drives. You will work side by side with the founders, turn a prototype that already drives into a vehicle that crosses real ground on its own, and set how our engineering team builds.

What you'll own

You own the autonomy stack end to end: how the vehicle knows where it is, what it sees, where it can drive, and how it gets there safely.

• Localization: fuse RTK GNSS, IMU, visual-inertial, and wheel odometry, and keep working when GNSS is jammed or spoofed.
• Perception and mapping: turn stereo depth (and lidar, if you make the case) into elevation and traversability maps. Tell rigid obstacles from grass the vehicle can push through, and find negative obstacles: ditches, holes, craters, drop-offs, and water.
• Planning: routes over terrain cost and mission constraints, local planning for a car-like vehicle with a ~2.5 m turning radius and coarse on/off drive, and recovery when it gets stuck.
• Obstacle avoidance and safety: stop-before-contact reflexes, e-stop, geofences, and safe behaviour when the radio link drops.
• Control: hold a path on slippery, uneven ground at low speed, and detect wheel slip.
• Operator side: teleoperation and 3D views that still work over thin radio links.
• Testing: a field-test program with logging and replay, simulation where it pays off, and metrics such as distance between interventions.
• Hardware: choose, mount and power sensors and compute; you will hold a soldering iron about as often as a keyboard.
• Team: set our engineering practices and help hire the next engineers.

What you start with

A prototype that already drives: you inherit a working stack, not a blank page.

• Vehicle: a converted electric off-road platform, 1.33 × 0.90 m, driven by wire through its own radio link, with no mechanical changes.
• Compute: an NVIDIA Jetson Orin Nano running ROS 2 Humble in Docker.
• Sensing: a Luxonis OAK-D Pro W stereo camera with a 9-axis IMU; RTK GNSS is next.
• Autonomy today: waypoint following from a browser console, and an obstacle guard that checks the path 29 times a second, 74 ms behind the camera. Nav2 is configured but not yet driving.
• Console: a web command centre with missions, live video and a 3D view of what the vehicle sees.
• Safety: a dead-man timeout, an e-stop, and a veto on forward drive when something is in the way.

You have

• 4+ years building autonomy for real outdoor robots or vehicles, and something you worked on drove itself outdoors or off-road.
• Depth in at least two of state estimation, perception and mapping, or planning and control, and working knowledge of the third.
• Off-road know-how: elevation or traversability mapping, negative obstacles, wheel slip, degraded GNSS.
• Strong C++ and Python, ROS 2 and Linux; at home on edge compute such as Jetson and CUDA.
• Hands-on hardware skills: sensors, wiring, power budgets, debugging in the field.
• A bias for field testing, and the judgment to ship a safe prototype in weeks, without a detailed spec.

Bonus

• Nav2 internals (Smac planners, regulated pure pursuit), grid_map or elevation_mapping.
• Visual-inertial odometry or SLAM (Basalt, OpenVINS, ORB-SLAM3, cuVSLAM), especially without GNSS.
• Lidar pipelines, or learned (self-supervised) traversability.
• Defense, agricultural, mining, planetary, or search-and-rescue robots, and time spent with end users in the field.
• Low-bandwidth teleoperation over mesh radios.

What we offer

• A founding role with meaningful equity, and a salary of [range].
• Ownership of the autonomy architecture from day one.
• Short loops: from an idea to a field test in days, with direct feedback from the people who will use the vehicle.
• San Francisco, with regular field testing.