Agentic Product Recommendation System
Conversational recommendation system combining multi-agent LLM pipelines with hybrid retrieval for e-commerce product discovery.
PythonFastAPIChromaDBPydanticRAG
Problem
Users couldn't express qualitative intent (taste, style, pairing) through traditional keyword or filter-based product search.
Solution
Combined structured SQL filtering with semantic vector search (ChromaDB + sentence-transformers), with schema-validated outputs and contradiction detection to prevent hallucination.
Impact
Reliable conversational product discovery with zero hallucinated attributes