Auterion · posted · checked
Senior Computer Vision Engineer
C++PythonOn-site · ZurichSenior
Who can apply
On-site · Zurich
Location: Zurich
Overview
We are looking for a Senior Computer Vision Engineer to build and deploy perception systems that run onboard autonomous aerial platforms. This is not offline research. Our algorithms run in real time on embedded computers, in the air, in conditions we do not control.
You will work across the full stack of the problem: classical computer vision and deep learning, camera and sensor selection, calibration, model training and evaluation, and optimization for constrained hardware. You will take your work to the field, fly it, look at the logs, and improve it. Our platforms serve both commercial and government customers across applications including inspection, mapping, search and rescue and security.
If you like hard real world problems and short feedback loops between an idea and a flight test, we want to talk with you.
Responsibilities and Duties
Design, implement and optimize onboard perception algorithms for real time detection, temporal association and state estimation of dynamic objects from a moving platform.
Train, retrain and evaluate neural networks for detection and classification, including dataset curation, labelling strategy and evaluation methodology.
Optimize models and pipelines for embedded inference, working to explicit latency, throughput and compute budgets.
Develop and maintain camera calibration and sensor synchronization procedures, including multi camera and multi sensor setups.
Combine visual measurements with inertial and platform state data using filtering and estimation techniques for robust performance under egomotion.
Plan and run field test campaigns. Take algorithms from bench to flight, work with pilots and integrators on site, and close the loop on real captured data.
Build repeatable evaluation methods so that performance and accuracy are measurable across every iteration rather than judged by eye.
Provide input into camera, lens, sensor and compute selection, and evaluate hardware options against real performance data.
Take full ownership of computer vision deliverables. Disambiguate requirements, drive alignment across teams, keep stakeholders informed and see work through to the customer.