In August 2023 I resigned from Meesho. I had a good title, good work, and the kind of trajectory people describe as "sorted." I traded it for a rented room, a stack of NCERT textbooks I'd last touched as a teenager, and a two-year attempt at the UPSC Civil Services Examination — the entrance exam to India's bureaucracy, taken by roughly a million people a year, cleared by fewer than one in five hundred. [1]
People assumed I'd had a breakdown, a bad manager, or a spiritual awakening. It was none of those. It was a spreadsheet feeling: I kept optimizing conversion funnels while the questions I actually cared about — why does a welfare scheme leak, why does a school fail, why is state capacity the binding constraint on everything in this country — sat in a tab I never opened. The civil services were, as far as I could tell, the only job where those questions were the job description.
So I studied. Polity, economy, geography, ethics, two thousand years of history, the entire apparatus of the Indian state. Ten hours a day, most days, for most of two years. I cleared the prelims of two state commissions — APPSC and TGPSC — and qualified for their mains. I cleared CDS, the defence services exam. And the one I actually wanted, the one the whole room was rented for, I did not clear.
What failure at scale feels like
Here is what nobody tells you about a 0.2% selection rate: it doesn't feel like a lottery while you're inside it. It feels like a meritocracy right up until the moment it doesn't. You can do everything correctly and still lose to variance — a bad question paper, an examiner's mood, the two marks that separate your rank from a different life. I had spent years at Meesho explaining to stakeholders that a single experiment result is noise and only the distribution matters. Then I bet two years on a sample size of one.
The exam taught me the syllabus. Not clearing it taught me the difference between a decision and an outcome — and that you can only ever be judged fairly on the first.
I don't say this as consolation. The decision was correct. Given what I knew, given what I wanted, given the asymmetry — a modest downside of two lost years against the upside of a life spent on problems I cared about — I would make the same call again at the same odds. That is not what losing gracefully sounds like; it is what expected value actually means when you apply it to your own life instead of a feature launch. [2]
What the room gave me
The strange thing is how directly those two years compound into product work. I came back knowing how the Indian state budgets, legislates, and fails — which turns out to be an excellent education in systems that have users, incentives, and feedback loops, just very slow ones. I came back with the ability to sit with a hard problem for months without the dopamine of a sprint demo. And I came back with calibration: after you've been outworked by variance in front of a million competitors, no quarterly review will ever frighten you again.
When I joined Dashverse in 2025, I was in some ways starting over — new industry, new stack, a resume with a two-year hole that recruiters scan like a wound. Within a year I owned the experimentation roadmap, had rebuilt the analytics layer, and had written the most consequential document of my career — a recommendation to kill a feature everyone wanted. I do not think the person who left Meesho writes that document. He didn't yet know how to lose something big and stand behind the reasoning anyway.
The room is empty now. Someone else is probably in it, ten hours a day, betting their own two years. I hope they clear it. And if they don't, I hope someone tells them what I'm telling you: the selection rate measures the seats, not the students. Judge the decision. Own the outcome. Keep the notes — you will be surprised where they compound.