{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/865e5a339cd346f7a3a9ea2f86e65fbb\" 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/865e5a339cd346f7a3a9ea2f86e65fbb-5b81d8be890bf2e4.gif","duration":377.579,"title":"Deterministic Music Recommendations Pipeline Demo","description":"In this Loom, I walk through the end to end pipeline I built for generating music recommendations. It includes input validation to normalize user profiles, ranking and scoring for deterministic results, evidence retrieval with RAG for grounded explanations, and a final presentation step. In the demo, the UI returns 3 top suggestions with a confidence level of 1, and different defaults produce different outputs based on the user profile. When I added a trailing whitespace, the system still worked but showed profile warnings. I also ran six test cases for normalization, guardrails, and workflow warnings, and all passed. If you want to run it yourself, I shared a GitHub link and use Streamlit for the UI."}