About
I have been building software since 2004, which means I shipped my first mobile applications onto BlackBerry and J2ME handsets three years before the iPhone existed. That timing turned out to matter more than I understood at the time: I learned early that being useful on a new platform beats waiting for it to settle.
Most of what I have done since has followed the same shape. I am currently with XY Retail, putting AI onto a fashion platform that already has stores. As CTO of Moneycontrol, part of Network18, I built a lending business inside a media company because the audience data was already there and nobody had asked what else it could do. At a pharmacy chain, twenty years of prescription records turned out to predict local disease outbreaks well enough to change how stores were stocked and where new ones opened. In the newsroom, agent systems researched and drafted at real volume — and every publish waited for an editor. None of that was on a roadmap when I arrived.
I have led engineering organisations that ship — the largest was 120 people across platform, mobile, data and security. The people who design the work should still be in the room while it is delivered. I look for operators who can write the constraint down: who owns this, what it may touch, what the approver sees, and what happens when it is wrong. I do not hire for a demo.
Under pressure I would rather pause a write than protect a timeline. A system that cannot wait for a person is not ready, no matter how good the model looks. I will not sign a percentage guarantee on work that is never actually handed over, and I will not open with a platform sale when the licences you already hold will do.
I also founded Onroll. I disclose that on every engagement, because advice from someone with a product to sell is only worth having if they will tell you when not to buy it.
I am currently with XY Retail, putting AI onto a fashion platform that already runs 1,300 stores in more than 50 countries. The work is the product, and it has to survive contact with merchandising, demand and a shop floor.
Network18 brought me back after the CTO tenure ended, for AI implementation through April 2026 — editorial and back office, on the licences the group already held. Training sits alongside client work rather than around a full-time job.
I write a column for Moneycontrol on AI and the future of work — the apprenticeship it quietly removed, the billing model it broke, and the office it was never designed for. Most of what I argue there comes from having had to ship the thing rather than forecast it.
I am a father to a daughter, which has taught me more about handing work over than any organisation did. You explain the constraint, you let them do it, and you resist the urge to take the keyboard back. It turns out that is the whole job.
I am tiresome about coffee. Beans, grind, water temperature, the lot — the same instinct that makes me ask what the approver sees, pointed at something that does not matter at all. Everyone around me has learned not to ask.
My favourite book is The Hitchhiker’s Guide to the Galaxy, and I do not think that is unrelated to the day job. It is a book about a spectacularly powerful computer that produces a technically correct answer nobody can use, because nobody did the work of understanding the question first. I have watched that happen in enterprises with a straight face. Douglas Adams got there in 1979.
I still read the papers the long way. The newsroom years left a habit of asking, before anything goes out, who owns this and what happens if it is wrong.
I read philosophy the way I read a specification — looking for the constraint someone hid in a clean sentence. Most bad systems fail the same way most bad arguments do: they answer a question nobody actually asked.
I like The Matrix for the same reason I like the Guide. It is a film about a demo that looks like the building. I have sat in rooms where everyone agreed the simulation was the work, and the hard part was saying so without sounding like you wanted a red pill for a living.
I spend my days writing agent skills — a file that says what a thing is for, when to use it, and when not to. It seemed only fair to publish my own. It is real: /skills.md serves it as plain text, so an agent reading this site on your behalf gets something it can actually parse. The other file is the gate I make you write before anyone builds: /gate.md.
# Rajeev Belani
## When to use this skill
## When NOT to use this skill
## Capabilities
## Constraints
## Known failure modes