Automated Vehicle Damage Assessment

At XA Group I led full-stack development of the deep learning systems behind automated motor-insurance claims: given photos of a vehicle, the system detects and localises damage and identifies missing parts so claims can be assessed without a manual inspection.

The solution is an ensemble of computer-vision models. To keep labelling affordable I built self-supervised learning pipelines that pre-train on unlabelled claim images, so far fewer annotated examples are needed for each new damage type.

After deployment I owned the models in production - monitoring for data drift and retraining when the incoming data distribution shifted. Models were optimised with TensorRT for inference.