Artificial Intelligence Engineer

About the role

FPT UAV designs and builds its own multi-platform UAV family (urban, field and cargo) under a technology-ownership mandate: flight controller, on-board computer, flight software and autonomy are developed in-house. The AI Engineer owns the on-board perception features – detection, tracking, segmentation and scene understanding – from dataset to a model running within the latency and power budget of the mission computer, and integrates them with the flight software used by public-security, Japanese and Korean customers.

Responsibilities

• Perception models: design, train and evaluate object detection, multi-object tracking and segmentation models for aerial imagery (EO and IR), covering people, vehicles, vessels and infrastructure in Vietnamese operating conditions.
• Edge deployment: optimise and deploy models on the on-board computer (Qualcomm / NVIDIA Jetson class SoCs) with TensorRT, SNPE or ONNX Runtime – quantisation, pruning, pipeline profiling – to meet real-time latency and thermal limits.
• Data pipeline: build the collection, labelling, augmentation and versioning pipeline from flight-test footage; define Vietnam-specific test sets and report metrics honestly on them (no cross-dataset averages).
• Integration: package perception as ROS 2 nodes / services on the mission computer, with clean interfaces to the PX4-based flight stack, gimbal, datalink and ground control software.
• Flight testing: take part in field flight tests, analyse logs, close the loop between field failures and model or pipeline improvements.
• Engineering quality: write maintainable C++/Python, unit and regression tests for models, and documentation; keep licence compliance of third-party models and datasets (e.g. AGPL-licensed detectors) under control before productisation.

Must-have requirements

• Bachelor's or Master's degree in Computer Science, Electrical Engineering, Applied Mathematics or related field.
• 3 to 5 years of hands-on computer-vision or deep-learning engineering experience with products in the field, not only research.
• Strong Python and solid C++; PyTorch (or TensorFlow) in daily use; OpenCV.
• Proven edge deployment: at least one model shipped on an embedded GPU / NPU (TensorRT, SNPE, OpenVINO, TFLite or ONNX Runtime), including quantisation and profiling.
• Comfortable on Linux: Docker, Git, CI, basic embedded-Linux debugging.
• Working English (technical documents, partner meetings); Vietnamese preferred for daily work.

Nice to have

• ROS 2 experience; integration with PX4 / MAVLink or other robotics middleware.
• Multi-object tracking (ByteTrack, SORT family), geo-localisation of targets from camera and IMU/GNSS, visual-inertial odometry or SLAM.
• Thermal / IR imagery, low-light or small-object detection, synthetic data (Isaac Sim, Gazebo, Unreal).
• Model licensing awareness and MLOps tooling (DVC, MLflow, Weights & Biases).
• UAV, defence, automotive ADAS or robotics background.

What we offer

• A greenfield in-house UAV programme where your models fly on real aircraft within months, not slide decks.
• Direct exposure to international partners and customers in Japan and Korea, and to national-scale programmes in Vietnam.
• Competitive package by seniority, FPT Corporation benefits, training budget and conference travel.
• Clear growth path: the AI team is planned to grow to six, with a technical-lead track open.