The One-Person Engineering Team

When I was running my gaming studio, I had three parallel development teams at peak — engineers working on three different games simultaneously. I was the CTO, and I loved it. Coordinating builds, reviewing architecture, shipping products. Technology was the business.

Then I started Cotton Candy Station. And the tech team became just me.

What Thirteen Years of Shipping Products Actually Builds

Before I get into what I built at Cotton Candy Station, I want to spend a moment on what the gaming years actually gave me — because it’s the foundation everything else stands on.

Building 40+ mobile games across 13 years is not just a volume number. It meant going through the full product lifecycle dozens of times: ideation, design documentation, prototyping, QA, deployment, live-ops, analytics, and eventually the hard call of deprecating something that wasn’t working. Our flagship, Draw N Guess Multiplayer, reached 12 million downloads. Getting there meant architecting real-time multiplayer systems on Google Cloud Platform — engineered for low latency, high availability, and concurrent global users — and owning every infrastructure decision along the way.

What that builds, over time, is a way of thinking. You learn to go from a business problem to a system design without needing someone to translate for you. You understand build-vs-buy trade-offs instinctively. You know what good looks like at every stage of the development cycle. That mental model doesn’t belong to gaming. It’s portable. And when I walked into a cotton candy factory with no engineering team, it was the only technical leverage I had.

The Question I Had to Answer

When I was simultaneously setting up the manufacturing unit, building the distribution network, and getting sales off the ground at Cotton Candy Station, a very practical question surfaced: where is the time going to come from to build the tech stack this business needs?

Because a manufacturing business without proper operational tooling is flying blind. You need order tracking, inventory visibility, production scheduling, sales performance data, distributor management, and revenue consolidation — at minimum. Without those systems, you’re making decisions on instinct and spreadsheets, and at scale that becomes expensive.

I already understood what needed to be built and how to build it. What I needed was execution velocity. That’s where AI changed everything.

ChatGPT became my earliest development partner. I used it to build our entire D2C store — WordPress + WooCommerce — and shipped 15+ custom plugins for store-specific workflows and checkout flows. Things that would have taken weeks with a freelancer came together in days. Then I moved to Claude for more complex work — internal tooling with real business logic, architecture decisions, debugging under pressure. For serious coding, nothing has matched it. That remains true today.

What I Actually Built

Once I understood that AI could dramatically compress development time for someone who already knew how to think in systems, I started building everything the business needed — one tool at a time.

The internal product suite I shipped independently: an Order Tracker, Inventory Management with AI-driven demand forecasting, Production Planning and scheduling, a Sales Incentive Dashboard, Sales Executive Performance Analytics, Distributor Management, Distributor Analytics, an Event Management tool, and a unified Business Revenue Dashboard that pulled live data from Amazon, our D2C store, and the distribution channel simultaneously — including ROAS connected directly to Meta Ads. Zoho Bigin CRM was integrated for pipeline and lead management. The entire infrastructure runs for under $10 a month in hosting.

Every one of these tools was designed, built, iterated, and maintained by a single person. When I compare that to what outsourcing this suite would have cost — the quotes, the timelines, the back-and-forth on requirements, the ongoing dependency on an external team for every feature change — it’s not a close comparison. I had full ownership, full control, and the ability to ship a new feature or fix a bug the same day the business need surfaced.

I also built the onboarding materials for the sales team — intro documents, videos, presentations, and demos — with AI doing a lot of the heavy lifting in putting them together. What started as basic training content evolved into a full sales playbook. Getting a non-technical sales team to adopt internal software and operate from a common playbook is its own challenge. Getting that right matters as much as the tools themselves.

Where This Is Going

The development paradigm has shifted faster than most people have caught up to.

I’ve been building with agentic AI for a while now — moving beyond single-prompt assistance into systems where AI agents handle multi-step workflows autonomously. I’m currently exploring OpenClaw, working on bringing my own digital operations into a more coherent agentic structure, and the pace of what’s becoming possible is genuinely exciting. Enterprise-grade software, built solo, is no longer a thought experiment. It’s happening.

For traditional businesses especially — manufacturing, distribution, FMCG — there’s a moment available right now that won’t stay open forever. The gap between companies that have woven AI into their operations and those that haven’t is widening every quarter. The barrier to entry for building serious internal tooling has collapsed. What required a development team two years ago requires a single person with domain knowledge and the right tools today.

The specific tech stack behind what I built — the architecture, the tools, the workflow — deserves its own post, and I’ll write it. But the broader point is this: if you’re running a business and you’re still treating technology as something you outsource and forget, you’re leaving compounding leverage on the table.

If you’re trying to figure out where to start with AI in your operations, I’m happy to talk.

— Sri