About
I build products for people I know, then check whether they actually work. By day I help enterprises buy AI-led delivery. Before that I shipped regulated fund-accounting software and did an MBA at IIT Kharagpur.
On paper that looks like three careers. In practice it is one question: is this worth building, and how do you know before you have built it?
In fund accounting the answer had to be provable. A wrong number cost a client money. A missed rule cost their trust. The MBA gave the question a language. Enterprise sales gave it the harshest test I have found: a buyer pays, or they do not.
Regulated finance: what being wrong costs
At Temenos I worked on a platform where being unforgiving about a wrong number was the requirement, not a slogan. I delivered the first client go-live on IBOR. HSBC. Forty-six funds into production, later past five hundred. The company gave me its GEM award for it.
What that work taught me is the discipline behind everything I build now. A number you cannot prove is a number you do not show. The gate on the dental ledger that stops an unverifiable figure reaching the dentist came straight out of that room.
Enterprise sales: what "worth it" means
I now sell AI-led delivery into enterprises. Six new logos, including a multilateral development institution and a clinical-trial technology provider. It is the most direct read I have on which AI capabilities get funded and which ones get a polite demo and no signature.
The same read has settled product arguments in my own work. While building EaseMySemester, the plan was to paywall the question bank. I argued the paywall belonged in front of the predictions instead. The objection was that we were already surfacing the numbers behind those predictions, so nobody would pay for them. That had it backwards. The bank is raw material. The prediction is the ranking already done for you. People pay for convenience, not insight.
It is easy to believe a demo. It is harder to believe a purchase order. My week is spent finding out the difference. That is information a product manager should have before asking anyone to build.
The proof
Somewhere in there I started building again, for people I know, where I could measure whether the thing worked. A ledger for a dental clinic. A question bank about fifty medical students use. A local search engine that runs on my own hardware.
Each one is a claim I can attach a number to. That is still the only way I can tell an idea from a product.