| 🚀 API Copilot — AI-Powered Developer Support for Payment APIs | • Built and shipped a production RAG platform using LangGraph, MCP, FastAPI, ChromaDB, OpenAI, and Claude to provide context-aware developer support for Payment API integrations.
- Implemented a 5-stage RAG pipeline: ingestion → chunking → embedding → semantic retrieval → grounded response generation.
- Integrated MCP-based intent-aware routing, observability, structured logging, Redis caching, and execution tracing.
- Designed a cross-provider DeepEval evaluation framework across OpenAI and Claude, achieving Faithfulness 1.0 and Relevancy scores of 0.906–1.0.🚀 Retirement Copilot — AI-Powered Retirement Intelligence Platform | • Built and deployed a retirement advisory platform using FastAPI, SSE streaming, RAG, OpenTelemetry.
- Implemented grounded pension guidance using semantic retrieval, vector search, and metadata-aware context enrichment across pension plans.
- Developed retirement forecasting workflows including corpus projection, SIP planning, pension estimation with AI guardrails.
- Integrated RAGAS and Phoenix observability, achieving a 28% improvement in groundedness while maintaining sub-$0.001 average query cost.🚀 Quasar Enterprise AI Delivery Swarm — Multi-Agent Governance Platform | • Architected and deployed an advisory-only multi-agent intelligence platform using Band.ai, FastAPI, Plotly, TimescaleDB, and AWS.
- Built live market data ingestion pipelines using Zerodha (MCX) and TwelveData (XAUUSD) with automated backfill, freshness monitoring, and persistent time-series storage.
- Developed multi-timeframe intelligence engines generating SMC (BOS, CHOCH, FVG), scenario analysis, hierarchy alignment, confidence attribution, and market memory.
- Designed specialist AI workflows covering Market Intelligence, Risk Governance, Delivery Planning, and Final Review, supported by governance evidence, audit trails.
- Submitted to the Band of Agents Hackathon 2026. |
