Computer Vision Engineer

Il y a 2 jours

Ghent Flemish Region, VLG, Belgique Augment Vision Temps plein
ABOUT UUG.
AI / AUGMENT VISION
We turn camera streams into useful, real-time signals for operations teams. As a Computer Vision Engineer, you work across the model lifecycle: from framing a problem and preparing data to evaluating, deploying and observing models in production. This is a full-time role with hybrid collaboration from Ghent, Belgium. WHAT YOU WILL DO - Build and improve detection, segmentation, classification and multi-object tracking pipelines. - Define datasets, evaluation sets and metrics that reflect real customer operating conditions. - Optimise inference for edge devices and GPU clusters, balancing accuracy, latency and cost. - Package and serve models through our production inference stack, including NVIDIA Triton where appropriate. - Investigate model failures using recordings, telemetry and user feedback, then turn findings into measurable improvements. - Collaborate with platform, workflow and frontend engineers so model output is understandable and actionable. WHAT YOU BRING - Hands-on experience building computer vision systems with Python and PyTorch or a comparable framework. - Understanding of model evaluation and representative test sets. - Comfort with video, OpenCV, image processing and production-inference trade-offs. - Ability to move between experiments and engineering work: tests, containers, APIs, profiling and operational debugging. - Clear communication of assumptions and results, with ownership and support from the team. HELPFUL EXPERIENCE — NOT ALL REQUIRED - YOLO-family models, tracking algorithms or segmentation workloads. - NVIDIA Triton, CUDA, TensorRT or GPU performance profiling. - MLOps, dataset versioning, model monitoring or automated evaluation. - Camera protocols, streaming media or edge-computing environments. - Go or C++ alongside Python. WHAT TO EXPECT Start by learning our pipelines and operating context. Take ownership of a measurable model or performance improvement, and grow toward end-to-end ownership from problem definition through deployment, monitoring and iteration. Share a relevant project or example of your work with your application.