Computer vision for component inspection

An Italian airline operator was running manual visual inspection of components during scheduled maintenance. The process was subject to operator fatigue: after hours of inspection, defect detection dropped and the risk of false negatives (missed defects) grew. We built a computer-vision system that assists the operator: the camera captures the image, the AI detection model (YOLO-based + custom classifier) highlights regions of interest with bounding boxes and probabilities, and the operator confirms or rejects. The system learns from corrections over time. Edge deploy (on the operator's tablet, no cloud) for latency and data residency. Full audit trail for every inspection, compliant with ENAC/EASA requirements. PoC duration: 8 weeks (real datasets need regulatory validation). Scale model: Tessaris run on client infrastructure, compliance handled directly.