Available now·Remote·Any timezone
Fourteen years in product teams, mentoring engineers, tech-leading delivery and owning the platform other teams build on. Seven products shipped end to end, three of them AI products with agents, retrieval and LLM pipelines in production. All still running. TypeScript, Node, React, Postgres.
14
years shipping software
8
companies, PayPal to Atom Learning
7
products designed, built and shipped
3
of those are AI products, all live
I've spent fourteen years inside product teams, the last five at Atom Learning, where I owned the design system the rest of engineering built on, mentored the people using it, tech-led delivery and interviewed the engineers we hired.
Alongside that I've shipped seven products of my own, three of them AI products: Rankahead.ai, Recly.io and Decklab AI. Some with cofounders, all of them end to end: design, frontend, backend, infrastructure, billing and the support inbox.
That is where the AI work stopped being a demo. Owning the model bill, the latency and the support inbox at once teaches you quickly which prompts are load-bearing, what a model does to inputs you never imagined, and how much of an AI product is really the plumbing around the model: the retrieval, the schemas, the retries, the queue behind it all.
It's the combination I'd like to be hired for. I can own an AI product end to end, model layer included, and I would much rather do it with a team than around one.
I've spent fourteen years building products in teams, and shipped seven of my own, three of them AI products that are live today: Rankahead.ai, Recly.io and Decklab AI. The AI work is production software with the whole apparatus around it, agents and tool use, retrieval-augmented generation, structured outputs and scheduled LLM pipelines, rather than notebooks or demos.
Queries four assistants on a schedule, reads the answers back to score how often a brand appears and which URLs get cited, then runs agents that turn each gap into answer-first content and publish it to the customer’s CMS. Customers bring their own keys, stored AES-256 encrypted.
Case studyModel layer
Detected clicks and scrolls become a step list, and each step is captioned by a model working from the surrounding UI context, so a raw screen recording arrives as a narrated walkthrough the author only has to edit.
Case studyModel layer
A one-line game concept is expanded into a full card set under a schema, so abilities and stats come back as data rather than prose. A balance pass validates the set and regenerates what fails it, and the art for every card is generated alongside.
Case studyModel layer
On each of these I owned the whole path: the model layer, the API, the database, the queue, the frontend, the billing and the support inbox. Some were built with cofounders, and the ones that were are better for it.
The full AI recordTracks and lifts how often AI assistants mention your brand.
TypeScript/Next.js/Node.js/LLM APIs
Bug reporting that captures the whole debugging context in one click.
TypeScript/React/Node.js/Session replay
Record, edit and share interactive product demos in minutes.
TypeScript/React/Node.js/Browser extension
Turns a card game concept into a playable prototype in minutes.
TypeScript/React/Node.js/LLM APIs
A booking page that builds trust and filters out low-intent leads.
TypeScript/React/Node.js/Calendar APIs
It is the least reliable thing in the system, so it gets treated like one: a schema at the boundary, retries and a fallback behind it, and a path that still does something useful when the model returns nothing worth using.
The work I am proudest of is the engineers who got better while I was there. Code review as teaching rather than gatekeeping, the context shared before it is asked for, and a hiring bar I would want applied to me.
A design system only pays off if the people on it move faster than they would without it. That makes adoption and API design the job, not Storybook coverage, and it is won by talking to the teams using it.
Model spend behaves exactly like a performance budget: invisible until it is a line item, and cheapest to defend every release rather than in a panic once a quarter. Prompt size, model choice and caching are architecture decisions.
Anything works on the input you chose while building it. What matters is the input a real user pastes in at 2am, and seven products where the support inbox was also mine is how I learned the difference.
Decisions, tradeoffs and the option that was rejected. Distributed across timezones for most of my career, and anything that only exists in a call effectively does not exist.
AI
Shipped in products that are live today.
Daily
What I reach for without thinking.
Fluent
Shipped to production, repeatedly.
Working knowledge
Enough to build and operate a product on it alone.
Aug 2021 – May 2026
Senior Frontend Developer
Nearly five years, the longest tenure, and the same years the side products were built.
London, UK · Remote
Sep 2018 – Aug 2021
Senior Frontend Developer
Zagreb, Croatia
Dec 2016 – Aug 2018
Senior Software Developer
Mumbai, India
Apr 2016 – Nov 2016
Senior Software Developer
Mumbai, India
Dec 2014 – Apr 2016
Cofounder
Mumbai, India
Apr 2014 – Apr 2015
Frontend Developer
Mumbai, India
Oct 2012 – Apr 2014
Software Developer
Mumbai, India
Jun 2011 – Aug 2011
Software Developer Intern
Chennai, India
Education
B.Tech, Information Technology
West Bengal University of Technology · 2008–2012
● Available now · Remote · Any timezone