In 2022 I left a Senior Engineering Manager role at Apple — fourteen years, two tours, a team I built from zero — to figure out business from scratch. My first product wasn't software. It was a velvet choker.
Roughly twice a month: build-in-public numbers from Velosite and the brands, career strategy for senior engineers, and how I actually work with AI agents. Subscribers get my agentic workflow checklist.
I'd spent seven years on Apple's Identity team, the last stretch in Hyderabad building the India engineering group from nothing to twenty engineers. It was good work on systems that hundreds of millions of people touch. And I could feel the shape of the next ten years: bigger team, same building, same city, same conversations. I wanted to know whether I could make something people would pay for when Apple's name wasn't on it. So I founded Stackable Creations Pvt Ltd and, deliberately, did not start with software.
Software founders skip the boring parts of business because software lets you. Physical products don't. If you sell a choker on Amazon you have to learn product selection, sourcing, listing copy, photography, keyword bidding, inventory, returns, and what a customer in Ohio actually wants — every layer, with real money on the line. I wanted a forcing function to learn business end to end before I built tools for other businesses. Jewellery was cheap to start, fast to iterate, and unforgiving on the details.
Search-data-driven selection. I'm an engineer; I treated the marketplace like a dataset. Amazon's search-term reports told me what people were typing and not finding. That is where the first products came from — not from taste, from demand.
Review moats. Once a listing crossed a few hundred honest reviews it became very hard to displace. Compounding matters more than any single launch. One brand now sits on 1,700+ reviews, and that number does more selling than I do.
Seasonal timing. October is our Super Bowl — costume, cosplay, Halloween. Learning the calendar, and stocking three months ahead of it, was worth more than any ad tweak.
Everything I know about marketplace advertising I learned the expensive way. My first campaigns burned budget on broad keywords that converted for nobody. I over-ordered a colour that didn't sell and under-ordered the one that did. I learned that “in stock” at a supplier and “on a shelf in a US warehouse” are separated by weeks and customs forms. None of this shows up in a Silicon Valley career, and all of it is business.
The most useful thing I learned wasn't operational. It was that the company was really two businesses. The people buying bold statement pieces for cosplay and alt fashion were not the people buying waterproof, non-tarnish pieces they'd never take off. Same category, different humans, different promises. So it became two brands: Stackable Creations for the self-expressive buyer, and Swogue for the woman who's always in motion — with you, wherever you go. Together they now do over $250k a year across the US, UK, Canada, Mexico, and Australia. Neither is a “jewellery brand”. Each is defined by the customer it serves.
Running two brands taught me how much storefront tooling gets in the way of small brands: the page builders that fight you, the marketing tools bolted on afterwards, the dozen logins. So I went back to what I'm best at — and started building Velosite, a multi-tenant storefront platform with an AI-assisted page builder, commerce, and a built-in marketing engine. This site runs on it. This post was authored by an AI agent through Velosite's own connector, under my review, which is roughly how I build everything now: agents in the loop for planning, implementation, tests, and deploys, with real engineering discipline around them.
More of this, in public: the numbers, the decisions, the mistakes. If you're an engineer thinking about your own leap — or a developer who wants to work with AI agents the way I do — the newsletter is where I write it up first.
