I need a proven AI routine that takes any of my paintings and instantly returns the most visually similar pieces from my collection. The focus is strictly on paintings and the only functionality I’m after is an accurate, fast artwork-similarity search—no tagging, no enhancement, just “show me what looks alike.” Because this is a production requirement, I can’t fund on-the-job experimentation. Come in with the method you already trust—whether that’s a CLIP or ResNet encoder feeding vectors into FAISS, Annoy, or another high-performance index—and wire it up so I can run it locally on a modest GPU or even CPU if feasible. Deliverables • A self-contained Python project (PyTorch or TensorFlow is fine) that: 1. Ingests a folder of painting images and builds/updates an index. 2. Accepts a query image and returns the top N similar paintings with distance scores. • A short README covering setup, commands, and expected folder structure. • A brief video or screenshots demonstrating the routine working on at least five sample queries. Acceptance criteria: the search returns visibly related paintings at the top of the list in under two seconds on a 1k-image test set. If you already have code that meets these points, I’m ready to review it and move quickly.