A YOLOv8 computer-vision system that grades pizzas against recipe specs from overhead photos —
ingredient/defect detection, a 100-point scoring engine, and a role-based multi-store dashboard. Runs ~1.5s/image on decent CPU;
deployed live across a franchise
1
14
A dedicated YOLOv8 detection model that finds hands/forearms in each frame and automatically blurs
identifying features (rings, tattoos, skin) before any image is stored — privacy-by-design for a GDPR-compliant
deployment.
1
28
A Gemini Flash pipeline that turns images and PDFs into clean structured data: text extraction with
bounding boxes, an LLM proofreading/QC pass, inpainting and automated typesetting. Benchmarked at F1 = 0.615 for ~€0.02
per document.