Notes

I prefer async work and as few meetings as possible. If I've hired someone, I've already seen something in them, so I give them time to settle in. Everyone comes from a different way of working, and I want them to push back on my calls too. When I have hard feedback, I say it directly, but always in private.

Early on I hired people only for their skills. I learned the hard way, at Vance and at Cube, that the kind of person someone is matters just as much. Most of the people I've worked with are friends now. One person I hired as an intern at Vance leads Android there today.

If you want someone who'll just agree with you, I'm probably not the right fit. I'll tell you what I think. Once we've decided, I'll commit to it.

I start by listening. Most of the first weeks go into five things.

1. How we work. How we release, how we build, and how product decisions get made.
2. The code. How it's designed, how tightly the parts depend on each other, how it performs, and which engineering numbers the team actually watches.
3. The team. How they think, how they work together, and how often they talk to users. Whether they use the app outside work, and what annoys them when they do.
4. The funnels. The engineering ones first, then the product ones.
5. The business. Why you hired me, what my numbers are, where growth is stuck, and what has to be done by Friday for the week to count.

Then I pick the bottleneck that matters most to the business and start there.

Many slow vendors, one portfolio

At Cube, showing someone's whole portfolio meant pulling data from many asset management companies, and their APIs were slow and failed often. We added ETags on the server and caching in the app, and people still sometimes saw stale or wrong numbers. We could only be as fresh as the vendors' APIs, so it became a product problem more than an engineering one. We designed the screen to show clearly which parts were up to date and which were still updating, and let people refresh any part of their portfolio on demand.

A KYC flow per country

At Vance, when we launched in the UK, we built a KYC flow for the UK's rules, then another flow for the next country. That was a mistake. KYC has a lot of moving parts and a lot of places where it can fail. Later, once we had our own licence and could run KYC ourselves, there were new ways to make it smoother, and every step still had to be encrypted and hold up to penetration testing. Remittance was crowded, and we wanted to change flows as we went without asking people to update the app. So we rebuilt KYC to be driven by the server while keeping the app native. The hard part was thinking long term about what compliance and legal would let us do and what they wouldn't.

Our own charts, for millions of people a day

At StockGro we moved the market module to charts we draw ourselves in Canvas, instead of an open-source library. We didn't want to be limited to standard bar and line charts, and we wanted to measure and tune the performance ourselves. With millions of daily active users, a mistake in something this central breaks the app for a lot of people at once. So we kept the old library, which we'd used for three or four years, and gave the new charts to one group of users at a time. Sentry and our logging showed us how the charts behaved on different phones and slow networks. We fixed what we found and widened the rollout. Within a month or two, everyone had them.

Some of my stints are short, so you'll probably ask. I've left when the direction changed and I didn't agree with it. I've never been let go. I'd rather not work than work on something I don't believe in.

I've been working at StockGro since early 2024.

Mobile felt creative to me because of its limits. You have less memory, a smaller screen and a device in someone's hand. Making something feel easy inside those limits is the fun part.

I started with Android because that's what most people in India use, so it was the biggest audience I could build for. Android being open also meant I could try almost anything. I've built a lot of iOS apps since, but Android is how I got here.

I didn't pick mobile because of the phone itself. I picked it because it's the closest I could get to the people using a product. I don't think the phone is the last device we'll use. My guess is voice, probably through glasses, within the next five years. Whoever is closest to users will figure out what works there first.