CV Engineer — Video Anonymisation Pipeline

We collect dashcam and bodycam footage for training autonomous vehicle and security models. Before we share footage with annotators or model training partners, we need to automatically blur faces and license plates. Needs to handle: varied lighting, partial occlusion, motion blur, multiple faces per frame. Must process 1080p video at 10fps minimum on our GPU server. Deliverables: Python pipeline, Docker container, evaluation report showing blur recall across test scenarios. Speed matters more than perfect recall — 97%+ is fine, we have a human QA step afterwards.