Senior Software Engineer at Netcore (Unbxd), Bengaluru. I build AI agent infrastructure and put it in front of real customers.
- MCP platform: I led the design and development of a multi-tenant Model Context Protocol server that lets AI agents operate Unbxd's e-commerce search and merchandising tools. It exposes 100+ tools, is live with 3 customers, and uses OAuth 2.1 so an agent cannot reach another customer's data. This is company code, so it is not public.
- LLM evaluation: an LLM-as-judge pipeline that scores AI-generated catalog data, simplified by measuring each stage before keeping it.
- Customer-facing AI: data modeling, integrations, and reusable agent skills that cut typical customer onboarding from 3 to 6 months to about 1 week.
Python, FastAPI, Node.js, TypeScript, React, Next.js, MCP, BigQuery, GCP
- partner-eval (Anvaya): a full-stack LLM evaluation app I built solo. It scores compatibility across 12 life domains through a pluggable provider layer (Claude, OpenAI, Gemini), computes the final rank deterministically, and ships with per-user token limits, an admin dashboard, tests, and CI. Live demo (invite-only sign-in).
- my-first-mcp: a learning project covering the three MCP primitives (tools, resources, prompts), with a Claude-powered CLI chatbot as the client.
- auction-website: a real-time multiplayer cricket auction app in Next.js and TypeScript, with admin, auctioneer, team owner, and spectator roles.
- react-search-sdk-demo: a minimal React app demonstrating a search SDK integration.
- FYP_GUI_real-time-sign-laguage-translation: a Django GUI for real-time sign language translation, from my final-year project. The related paper is published at IEEE.
- A multi-agent fact-checking system.