we are looking for a Computer Vision & Autonomy Engineer.
What you'll own
Design and own multi-sensor fusion pipelines – camera, radar, GNSS/INS, and altimetry – delivering robust, high-rate state estimation through GNSS-degraded urban environments.
Build the landing-site perception system: real-time detection, geometric assessment, and obstruction monitoring of vertiports and contingency sites during steep, noise-optimal approaches.
Train, quantize, and deploy neural perception models on embedded flight compute, owning latency, memory, and power budgets down to the millisecond.
Develop perception for detect-and-avoid: tracking non-cooperative traffic – small UAS, birds, general aviation – at ranges that preserve quiet, gentle avoidance maneuvers.
Stand up the data engine: flight-test data collection, ground-truth generation, dataset curation, and automated regression of model performance across releases.
Validate the perception stack in high-fidelity simulation, sensor replay, and hardware-in-the-loop benches before any algorithm touches the vehicle.
Architect run-time monitoring and assurance around learned components, producing the evidence structure a certification authority can eventually accept.
Support flight test campaigns end to end: integrate and calibrate sensors, analyze logs, and turn every anomaly into a fix or a documented limit within days.
What you'll own
Design and own multi-sensor fusion pipelines – camera, radar, GNSS/INS, and altimetry – delivering robust, high-rate state estimation through GNSS-degraded urban environments.
Build the landing-site perception system: real-time detection, geometric assessment, and obstruction monitoring of vertiports and contingency sites during steep, noise-optimal approaches.
Train, quantize, and deploy neural perception models on embedded flight compute, owning latency, memory, and power budgets down to the millisecond.
Develop perception for detect-and-avoid: tracking non-cooperative traffic – small UAS, birds, general aviation – at ranges that preserve quiet, gentle avoidance maneuvers.
Stand up the data engine: flight-test data collection, ground-truth generation, dataset curation, and automated regression of model performance across releases.
Validate the perception stack in high-fidelity simulation, sensor replay, and hardware-in-the-loop benches before any algorithm touches the vehicle.
Architect run-time monitoring and assurance around learned components, producing the evidence structure a certification authority can eventually accept.
Support flight test campaigns end to end: integrate and calibrate sensors, analyze logs, and turn every anomaly into a fix or a documented limit within days.
Requirements:
M.Sc. in computer science, electrical engineering, or robotics – or a B.Sc. with equivalent demonstrated depth in perception systems.
4+ years building computer vision or perception software that shipped on real autonomous platforms – aerial, automotive, or robotic.
Strong modern C++ (C++17 or later) for real-time perception pipelines, plus fluent Python for training, tooling, and analysis.
Deep grounding in state estimation and sensor fusion: Kalman filtering variants, factor-graph optimization, or visual-inertial odometry in production.
Solid geometric computer vision fundamentals – multi-camera calibration, epipolar geometry, structure from motion – not just learned end-to-end models.
Hands-on deep learning for perception in PyTorch, including optimization and deployment to embedded accelerators via TensorRT, ONNX Runtime, or equivalent.
Demonstrated rigor in validating perception systems: ground-truth methodology, failure-mode characterization, and metrics you would defend at a safety review.
Field experience: you have debugged your own algorithms on real hardware outdoors, where the sun, the dust, and the multipath do not care about your test set.
M.Sc. in computer science, electrical engineering, or robotics – or a B.Sc. with equivalent demonstrated depth in perception systems.
4+ years building computer vision or perception software that shipped on real autonomous platforms – aerial, automotive, or robotic.
Strong modern C++ (C++17 or later) for real-time perception pipelines, plus fluent Python for training, tooling, and analysis.
Deep grounding in state estimation and sensor fusion: Kalman filtering variants, factor-graph optimization, or visual-inertial odometry in production.
Solid geometric computer vision fundamentals – multi-camera calibration, epipolar geometry, structure from motion – not just learned end-to-end models.
Hands-on deep learning for perception in PyTorch, including optimization and deployment to embedded accelerators via TensorRT, ONNX Runtime, or equivalent.
Demonstrated rigor in validating perception systems: ground-truth methodology, failure-mode characterization, and metrics you would defend at a safety review.
Field experience: you have debugged your own algorithms on real hardware outdoors, where the sun, the dust, and the multipath do not care about your test set.
This position is open to all candidates.
