Technology partner
The engineer who stays in the room.
Fifteen years of software. A support desk, a startup that failed, an agency I co-founded, and a platform fifteen million people use. Each one paid for the next.
support desk to platform architect
users on Simplified, the platform I architect
restaurants paid for CeleryHQ — the last until 2026
continents, worked from Pune
on a support desk — worst hiring market in a generation
people use what I architected at Simplified
customers kept paying after we stopped trying
New Zealand to California — different clocks, languages, cultures. At home in all of them.
Seven chapters · 2009 — present
Nothing went to waste.
Every chapter below paid for the one after it. That is the whole argument.
CHAPTER 01 · 2009 · THE SUPPORT DESK
Where nobody's portfolio starts.
My first tech job was answering support calls. Fifteen years later, it's still the job: someone's thing is broken, they're stressed, and they need somebody calm who listens before fixing. The systems got bigger. The job never changed.
CHAPTER 02 · 2010 · THE SOLO BUILD
One junior. One whole product.
My first engineering job handed me an entire platform to build for yoga studios — schema, backend, frontend, deployment. I shipped it alone, learning as I went, and found the thing that has defined every job since: I can carry a product end to end. By the time I left, I was helping hire and mentor the Python team growing around me.
CHAPTER 03 · 2013 · FIRST ENGINEER
Employee number one of a team that didn't exist.
A Pune software firm hired me as their first Python engineer. Three years later there was a department of eight, a string of enterprise Django projects delivered on my watch, and my fingerprints on a marketing-automation platform later part of a $4.75 billion acquisition. The pattern was forming: get hired before the thing exists, leave behind a thing that works.
CHAPTER 04 · 2016–2018 · THE STARTUP
The garage years.
By day, I led frontend at the Pune center of a US retail giant — introducing Angular to a floor that had never shipped a single-page app and delivering their first one. The nights belonged to something better.
Three of us — friends, enthusiasts, more conviction than sleep — spent those nights building CeleryHQ, a cloud point-of-sale for Indian restaurants. I built the platform end to end. Real product, real customers, real revenue. The journey started well — well enough to teach us the hard lesson gently: it never became a company.
CeleryHQ didn't teach me how to fail. It taught me the difference between building a product and building a company — and that the difference is commitment, not code.
A note to self, around 2018CHAPTER 05 · 2018 · THE AGENCY
This time, all in.
The lesson didn't stay theoretical for long. In 2018 I did the thing we never did for CeleryHQ: I committed. Gave up the day job and co-founded Vellestudio — an IT consulting and digital marketing agency — with one of my two partners from the garage nights.
Under that umbrella we've built for an automotive-AI startup in the US, an options-trading platform, an education non-profit that's now been with us ten years — different domains, five continents. And the number I'm proudest of isn't on any chart: every client we've ever had arrived by word of mouth. No sales team, no ads, no cold emails. The work sells the work.
And the best referral of all was still on its way.
CHAPTER 06 · 2021 · THE CALL BACK
The best referral.
Then the third of us called. He was building something quietly, and he didn't run a search. I joined. It became Simplified.
What followed were five years inside one product's evolution. It began as a tool for making content, grew into publishing, then into generative AI, and now into workflows and agents. Each turn came from the market, not from us — and each one asked the platform to become something it hadn't been built for. My job was to make sure it could, without the customers ever feeling it.
The infrastructure has been mine to run from the start — no DevOps team, no platform team. That included moving production from AWS to GCP while millions of people were using it: a year's worth of learning done in the one environment where mistakes are expensive. Kubernetes, pipelines, cost control, and the phone that rings at night — all the same person.
Pricing moved constantly, the way it does in a startup. Old plans, new plans, and the one nobody had thought of yet, all live at once. The business decided what to charge; I spoke for the code — what it would cost, what it would break, and what we could do instead.
We stayed a monolith long after it stopped being fashionable. For a small team that was speed: one codebase, one deploy, features shipping while others were still wiring services to each other. We split off only what scale genuinely demanded. The strain in between was mine to carry — the honest price of choosing speed, and I'd pay it again.
When investors looked closely at the technology, I was the one who answered.
Fifteen million people use it. I still know how most of it works, line by line.
CHAPTER 07 · NOW · PRESENT TENSE
Specs, judgment, and a shuttlecock.
The way we build changed again, recently and completely. The team writes specs; AI writes most of the code; the work that matters has moved to deciding what should exist and judging what comes back. It's the largest shift in how software gets made since I started, and I'd rather be early to it than spend my forties defending the old way.
What hasn't changed: I still read the code. Fifteen years in, I've yet to find a substitute for knowing exactly how the thing works.
Most mornings, badminton. After a career of decade-long games, there's a particular pleasure in a point that ends in seconds.
DOMAINS
Every industry teaches you something the last one couldn't.
Restaurants taught me billing. Trading taught me latency. A non-profit taught me frugality — every rupee spent on infrastructure was a rupee not spent on teachers. Marketing AI taught me scale. The domain changes; the discipline doesn't.
GEOGRAPHY
One desk, most of the world's clocks.
New Zealand mornings, European afternoons, California nights — six countries over fifteen years, and a working day shaped around whoever needed me. Different timezones, different languages, different ways of saying no. Reading the room across a video call in an accent that isn't yours is a skill, and it took years; these days I don't think about it at all.
TOOLKIT
Things I've actually run in production.
One rule for this list: everything on it has been live, under real traffic, on my watch. Ask me about something that isn't here and I'll tell you plainly that I haven't shipped it.
WORK WITH ME · FOUR WAYS IN
Four ways to work with me.
Some people need three days. Some need three years. Both are fine — start where you are.
“Look at this and tell me what’s wrong.”
“Be the senior person we don’t have.”
“We need it built.”
“Come and do this with us.”
Or a game of badminton, if you’re in Pune. No agenda, nothing to sign. I like meeting people who build things.
For your company, college or meetup — architecture, AI in production, or what a failed startup teaches you. Ask and I’ll come.
Not technical work: vellepulse, a Vellestudio vertical, runs social growth marketing for health practitioners. If that’s you, this is your way in.
Not sure which one? Just say what’s going on. I’ll tell you honestly whether I’m the right person — or whether you don’t need anyone yet.
Book a call — 30 minutes, no pitch →
THE PUBLIC RECORD
Don't take my word for it.
A biography is a story its author gets to tell. These are the parts strangers can verify without me.
One handle everywhere: @pyarun — LinkedIn, GitHub, GitLab, Stack Overflow.
— chapter eight is unwritten —
Arun Mittal builds software in Pune, India. He co-founded Vellestudio, architected Simplified to fifteen million users, and works with founders on the technology calls that are expensive to reverse — as an advisor, or as the person writing the code.