I build AI systems that
think, decide,
and act.

Hours of manual data entry per invoice. Humans making errors copying values into databases with no audit trail.

Built a LangGraph multi-agent pipeline: one agent extracts, one validates, one flags uncertainty and pauses for human review before committing to the database.

Fully automated flow with an auditable trail. What took hours now runs unattended — with a human-in-the-loop safety net for edge cases.

Research that takes an hour of searching, reading, and synthesizing — nobody should do this manually when an agent can do it in seconds.

Deployed a LangChain + RAG pipeline: queries the live web, validates source credibility, chunks and embeds results, then synthesizes a grounded summary.

Live demo deployed on Render. Query → validated summary in seconds, with source traceability built in.

Flight search UIs are complex and time-consuming. People know what they want in plain English — they just can't express it to a form.

Built an LLM agent with LangGraph that parses natural intent ("cheapest flight to Mumbai next Friday"), resolves ambiguity through conversation, and calls live flight APIs with the right parameters.

Conversational flight search with real options. The agent handles follow-up clarifications and maintains booking context across turns.

Finding a trusted electrician or plumber locally is fragmented and unreliable. No centralized, verified platform existed for hyperlocal service booking.

Built the full stack: Spring Boot backend with JWT auth, React frontend, MySQL for data, Google Maps for location discovery, real-time WebSocket chat, and admin dashboards for dispute resolution.

Live on Vercel. Providers self-onboard with skills and location. Customers see nearby options on a live map and book with real time slots.