Computer Vision Engineer — Aerial Perception & Target Analysis
Location: Berlin (Onsite)
Job Type: Permanent
Role Summary
We are seeking a skilled Computer Vision / AI Engineer to develop and deploy robust AI-powered pipelines for the detection, classification, and real-time tracking of UAS targets. Working across IR/thermal and visible-spectrum imagery, you will design models and inference workflows that perform reliably under challenging conditions — low contrast, clutter, occlusion, and variable visibility. Your work will directly underpin system awareness, target identification, and track continuity. Onsite work in Berlin and NATO member state citizenship are mandatory.
Key Responsibilities
Detection & Classification: Design, train, and optimise AI/ML models to detect UAS targets and classify types/behaviours from IR/thermal and visible imagery
Real-Time Tracking: Integrate detection outputs into multi-target tracking pipelines, ensuring stable association, low-latency inference, and consistent identity maintenance
Model Optimisation: Adapt and compress models for edge deployment; implement inference pipelines using TensorRT, CUDA, and efficient architectures to balance accuracy and speed on NVIDIA hardware
Robustness & Adaptation: Improve performance across weather, lighting, occlusion, and sensor noise; implement data augmentation and domain adaptation to bridge simulation-to-reality gaps
Sensor Alignment: Work with thermal, radar, and laser teams to align imagery and detection outputs, enabling multi-sensor fusion and enhanced reliability
Evaluation & Iteration: Build rigorous evaluation frameworks; analyse field data and failure cases; drive continuous improvement in precision, recall, false-positive suppression, and latency
Documentation & Deployment: Maintain clear records of model architectures, training data, performance characteristics, and deployment constraints
Required Qualifications
5+ years’ experience in computer vision, deep learning, or real-time perception systems
Proven track record in object detection, classification, and multi-object tracking
Deep understanding of CNNs, transformer architectures, loss functions, and training methodologies
Proficiency with PyTorch or TensorFlow; practical experience in model compression, pruning, and quantization
Strong Python; working knowledge of C++ for production-grade inference deployment
Experience optimising and deploying vision/AI pipelines on edge platforms — ideally NVIDIA Jetson/Tegra
Familiarity with TensorRT, CUDA, and efficient inference practices
Ability to work cross-functionally with hardware, radar, and tracking engineers
Availability for onsite work in Berlin; NATO member state citizenship
Experience with IR / thermal imagery or low-contrast, low-resolution target detection
Multi-sensor or fusion perception experience (vision + radar/RF + laser)
Hands-on work with tracking-by-detection, temporal association, or re-identification
Familiarity with operational constraints: latency, SWaP, environmental robustness
Publication or project experience in aerial perception, small-target detection, or defence-related vision
#J-18808-Ljbffr
