Recommendation Agent
Personal
Aug 2025 - Current (1 year 3 months)
I engineered an AI-driven semantic recommendation engine and its cross-platform Flutter frontend, designing a hybrid multi-database architecture to match natural language user intents with enriched local business data. On the backend, I built data pipelines to ingest foundational profiles via the Google Places API, which I then processed using Google Gemini to extract abstract features (such as lighting, music type, and specific "vibes") and generate high-dimensional vector embeddings. To serve these insights efficiently, I decoupled the data layer: I implemented Pinecone for low-latency vector search and cosine similarity matching, utilized Supabase (PostgreSQL) to guarantee consistent relational storage for user authentication and transac