I build production-ready AI agents and LLM-powered backend services — systems that reason, use tools, and return structured, reliable outputs. My work is grounded in real projects: Artha, a multi-agent stock market assistant built with LangChain, live financial data, forecasting models, and news search, is a working example of the kind of system I deliver.

Agent Architecture Agents designed around a clear tool-use loop — the model decides what to invoke, when to invoke it, and how to structure the response. Whether you need a single ReAct agent or a coordinated multi-agent pipeline, the architecture is built to be debuggable, extensible, and production-stable. Implemented in LangChain with full control over memory, context, and tool definitions.

Tool & Data Integration Agents are only as useful as the tools they can reach. I integrate external APIs, document parsers, web search (Tavily, NewsAPI), vector stores, and custom analytical tools into the agent's decision loop — so it can act on real-world data, not just its training weights. RAG pipelines for document-grounded question answering are included where needed.

Backend & API Delivery The agent is served through a clean, documented API built in FastAPI or Flask — structured responses, optional chart or analytical data payloads, and a backend organized for easy handoff or future expansion. You get a system your team can maintain, not a notebook experiment.