Members-Only
Recent Talks & Demos are for members only
You must be an AI Tinkerers active member to view these talks and demos.
RAG for diet plans: teaching an AI to actually remember its user
Learn how to build a personalized AI nutrition planner using RAG. This talk details a retrieval system that remembers user history for accurate meal plan generation.
myfitplan is an AI-powered nutrition platform that generates personalized meal plans from each user’s goals, routine, restrictions, and progress over time. In this demo I walk through the RAG pipeline that turns scattered user context (preferences, past plans, weekly feedback, dietary restrictions) into a coherent memory the model can actually use, built on Supabase for retrieval and Langfuse for prompt versioning and full-pipeline tracing. Live, I will generate a plan for a real user profile and show exactly what gets retrieved, what gets injected, and why the output looks the way it does.