Neil Dasgupta

I design and ship production AI agents that turn messy, scattered data into something you can actually use structured, visualized, and queryable in plain language. If your company wants its first real AI agent (not a demo that dies in a slide deck), I take it from idea to deployed in about four weeks.

"Family Butler": I built an agent that connects to a Gmail inbox, parses two years of order-confirmation emails into a structured Postgres database, charts the spending, and exposes a conversational "butler" (powered by the Anthropic API) you can interrogate in natural language "how much did I spend on X last quarter?" It's a passphrase-protected web dashboard, deployed on Vercel (Next.js) with an automated cron sync that keeps the data fresh. Email in, structured insight out, conversation on top.

Fixed-scope, outcome first. You get a working agent you own and can maintain, clear documentation, and a handoff not a black box or an open ended day rate that drifts. I respond within 24h and keep you in the loop with demos, not jargon.