Consumer platformFrom 2021

Building a consumer platform from nothing

I joined as the founding engineer before there was a product, built the platform the company launched on, and stayed with it for years afterward while the business tripled year over year.

CompanyA B2C company
RoleFounding engineer, full time
3xGrowthyear over year after launch
1Engineers on the platform
On timeLaunched
Where it started

There was an idea, a small team, and no software. Everything the company planned to sell depended on a product that did not exist yet, and the runway assumed a launch date rather than a discovery phase.

That meant the first architectural decisions had to hold up for years, before anyone knew what the business would actually become. If they were wrong, the cost would arrive much later as a rewrite the company would struggle to afford.

How I built it

I built for the company we could reasonably become rather than the one we were that week. In practice that mostly meant keeping the data model honest and refusing to hardcode assumptions that were really guesses about the market.

Everything shipped in small pieces against something the team could click. A founder will skim a spec and miss the problem, then find it in ten seconds on a working screen, so I gave them working screens.

What shipped
  • The consumer-facing product, end to end
  • The backend, data model, and API it ran on
  • Authentication, accounts, and the payment flow
  • The internal tooling the team used to run the business day to day
  • Deployment, environments, and monitoring
Decisions

Three that mattered most

01

I was the whole engineering team

I owned the product, the backend, the infrastructure, and the internal tools at the same time. That constraint shaped the architecture more than anything else, because everything had to keep running without me watching it.

ArchitectureOwnership
02

A data model that survived the pivot

The product the company sells today is not the one we launched with. The schema absorbed that change without a migration project, because the early modelling described what things were rather than how the first version happened to display them.

Data modeling
03

Launched on a fixed date

Launch was tied to money, so the date could not move and the scope had to. We decided early and in writing what would not be in version one, which is what made the date achievable.

Delivery
Where it ended up

The platform launched on schedule and became the product the company sold. Revenue tripled year over year in the seasons that followed, on software I had built and was still the only engineer maintaining.

This is the work I point at when someone asks what I actually do. Not a feature added to an existing system, but the system itself, from the first commit through to the part customers pay for.

Built with
Frontend
NextJSReactJSTypeScript
Backend
NodePostgreSQL
Infrastructure
SentryPostHog
Services
Contentful