Production AI agents, and a faster CRM, for a VIP matchmaking company
AI agents that chat with clients, coordinate dates and book calls, plus the performance work that keeps the CRM behind them fast.
AI product engineer · Chennai, India · open for contract work
I build AI products end to end, and ship them.
LLM agents inside your product, whole products from idea to production, and the performance work that keeps them fast. 9+ years of shipping software, now focused on AI that takes real actions for real users.
- documents examined by a production booking query I rewrote
- 2.18M → 1,684
- per window for a sign-language model running offline in the browser
- 12.5 ms
- in two weeks for an AI-assistant builder I lead
- 5 releases
Teams I've built for
- Blush
- Toldo
- MessageBlue
- KERN & SOHN
- ZS Associates
- Genyus
- ToothHQ
What I can do for you
Four ways to work together
Each one starts with a call and a written proposal with a fixed price, so you know the cost before any work begins.
2–4 weeks
AI agent sprint
A working LLM agent inside your product, taking real actions safely.
What's included 4–8 weeksProduct build
From an idea or a prototype to a production v1.
What's included 1–2 weeksPerformance and reliability rescue
Slow queries, timeouts and crashes found, fixed and measured.
What's included MonthlyFractional senior engineer
A senior engineer on your team, part-time, month to month.
What's includedSelected work
Shipped, live, and measured
Production AI for a matchmaking company, a product built from scratch, and my own products you can open right now.
Toldo: build an AI assistant by chatting
A no-code AI-assistant builder that runs entirely on Cloudflare, from the first commit to paid production.
Sanket: Indian Sign Language recognition that never uploads your video
A 4.6 MB model that reads signs from a webcam at 12.5 ms per window, fully offline in the browser.
The agent layer of an iMessage API for AI agents
Retrieval over each customer's knowledge, memory, voice notes, and tool integrations that can't write without permission.
Ooru: three-word addresses and a citizen power-cut map for Chennai
A three-word address for every 12.5 m of Chennai, and a live map of power cuts that resists abuse without asking anyone to log in.
Deck: a presentation engine with a live audience room
MDX slides with a live room on Cloudflare: presenter sync, polls, quizzes and drawing from the audience's phones, with nothing to install.
Why work with me
What you can count on
Production, not demos
Everything I show here runs for real users. I design for the failure cases: wrong dates from a model, duplicate bookings, a query that times out on the busiest day.
Fast, and verified
I build with AI coding agents in parallel, so one engineer ships at the pace of a small team. Nothing ships on an agent's word: every change passes typecheck, tests and a check in a real browser.
Yours from day one
Code in your repository, running in your accounts, with a live preview URL from the first week and a demo every week after. No lock-in, and nothing to hand over at the end because you had it all along.
How it works
From brief to handover
- 01
Brief
Tell me what you need through the form. I reply within one working day.
- 02
Call
A 30-minute call to understand the problem, the users, the constraints and what done looks like.
- 03
Proposal
A written proposal within two working days: scope, milestones, timeline and a fixed price per milestone.
- 04
Build
Work in the open: a live preview URL, a demo every week, and every change through typecheck, tests and a real-browser check.
- 05
Handover
The code in your repository, deployed in your accounts, with docs and a walkthrough. It's yours.
Have something to build, fix or speed up?
Send a short brief. I reply within one working day, and you get a written proposal with a fixed price within two working days of our call.