2025·Featured·AI / RAG / Full Stack
CardMax AI
RAG-powered credit card intelligence for India — sub-second reward optimization.
Overview
CardMax is an end-to-end RAG product I built from scratch. It ingests bank PDFs, JSON, and SQL into a vector store and uses Google Gemini 2.5 Flash to answer context-aware reward optimization questions across HDFC, SBI, Axis, and ICICI in under a second.
Highlights
- Pipeline ingests PDFs, JSON, and SQL into normalized vector embeddings.
- Gemini 2.5 Flash for sub-second, context-aware reasoning at low cost.
- Coverage across 4 major Indian banks and dozens of card variants.
- Shipped on Vercel with PostgreSQL — single-developer end-to-end.
Tech stack
- Python
- React
- PostgreSQL
- RAG
- LangChain
- Gemini API
- Vector DB
- Vercel