{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/6b08d60d1b8a4c2786b2bb998fd7e959\" frameborder=\"0\" width=\"1728\" height=\"1296\" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>","height":1296,"width":1728,"provider_name":"Loom","provider_url":"https://www.loom.com","thumbnail_height":1296,"thumbnail_width":1728,"thumbnail_url":"https://cdn.loom.com/sessions/thumbnails/6b08d60d1b8a4c2786b2bb998fd7e959-3cb643cacd4487e3.gif","duration":318.546,"title":"Building a Reliable Retrieval AI System","description":"In this Loom, I explain how I turned my Week 7 Music Recommender into a full Applied AI system with retrieval, validation, and logging. The system retrieves relevant documents from a small knowledge base, generates answers from that context, then a validator assigns a confidence score and checks whether sources were found. It also logs every interaction as JSON files for debugging and evaluation. I tested in-scope questions like prompt engineering and retrieval augmented generation successfully, and out-of-scope overfitting triggers a fallback instead of a hallucinated answer. I did not ask viewers to take any action."}