Artificial Vision Engineer – Deep Learning & Edge Navigation

Your mission will be to design, optimize, and deploy Deep Learning algorithms applied to autonomous navigation in UAVs operating in GNSS-denied environments, contributing to the development of Terrain Matching systems capable of matching, in real time, aerial images captured by the drone with satellite reference maps.

The new team member will be responsible for taking artificial vision models from the development phase through to efficient execution on embedded hardware, working in an Edge Computing environment where performance, robustness, and clean integration with the existing architecture are critical factors.

Your contribution will be key to feeding the navigation system already developed internally, integrating with geometric modules such as RANSAC and PnP, and ensuring that the implemented solutions are efficient, traceable, and compatible with the company's high-integrity software standards.

Ultimately, your mission will be to turn advanced artificial vision capabilities into real, deployable, and reliable solutions for autonomous aircraft that must operate without relying on GNSS signals.

• Development and Optimization of the AI Pipeline: Implement, train, and optimize Deep Learning models for feature extraction and description in cross-domain images - camera vs. satellite - seeking maximum performance at the Edge.
• Software Integration and Connectivity: Connect the outputs of the AI model with the existing traditional geometric backend - RANSAC, PnP - ensuring efficient, low-latency communication.
• Robust Development with No Regressions: Integrate the code into high-integrity C++ modules, respecting the principles of the current embedded architecture - efficient memory management, absence of blocking operations - to ensure that new functionalities coexist seamlessly with the system's quality standards.
• Validation in Simulation and HIL: Validate the behavior of the Terrain Matching system in closed loop using our photorealistic virtual environments and Hardware-in-the-Loop systems before real flight testing.

• Degree in Computer Engineering, Telecommunications, Electronics, Mathematics, Robotics, or similar.
• Practical experience in training, evaluating, and exporting Deep Learning models applied to vision - PyTorch, TensorFlow.
• Solid programming skills in C++ - object-oriented programming, data structures - with the ability to integrate autonomously into a production software repository.
• Experience or familiarity with deploying networks in optimization runtimes for embedded hardware - TensorRT, ONNX Runtime.
• Fluent use of standard computer vision libraries - OpenCV.

• Experience or projects involving local descriptors, both traditional and learned: ORB, SIFT, SuperPoint, SuperGlue, etc.
• Familiarity with embedded platforms - NVIDIA Jetson or other ARM architectures - and Linux environments.
• Ability to work under real-time system constraints - latency control, buffer optimization.
• Knowledge of or interest in clean and critical code standards - structured coding style, MISRA C++, JSF++.