About
A software company built around one idea
That the infrastructure every application needs should be built once, properly, by people who will have to live with it — and then shared.
The story
Aftian started with a marketplace. Building it meant building the things a marketplace needs: a way to sign in, a way to take mobile-money payments, a way to hold a wallet balance and prove it was right, a way to find a place in a city where addresses are landmarks rather than street numbers.
Then came a retail system, and it needed the same four things. Then an ISP platform, and it needed them too — plus recurring billing. The obvious move at that point is to copy the code across. We did the other thing: we pulled those parts out into services of their own, gave each one a single job, and made every product a client of them.
That decision is the company. It is why a fourth product costs a fraction of the first, why improving the mobile-money integration improves four applications at once, and why no balance anywhere in our systems is a number that somebody incremented and hoped about.
We build from Cameroon, for the way business actually works here: mobile money before cards, prices in XAF, places described the way people describe them, and applications that have to work on a connection that comes and goes.
How we work
What we hold to
Build it once
Nine services exist so that four products do not each need their own identity, payments and accounting. A fix to the ledger is a fix for every product at the same moment - and a bug there is one bug, found once.
Money is accounted for, not stored
No balance on this platform is a number someone increments. Balances are the sum of double-entry transactions, which means every figure can be traced to the events that made it.
Built for how payment works here
Mobile money first, not a card form with mobile money added later. Prices in XAF. A gazetteer that understands landmarks rather than street numbers.
Failure should be loud
The expensive bugs in a distributed system are the silent ones - a service answering 200 while doing nothing. Credentials are tested against a wrong-key control, not assumed, and a misconfigured service refuses rather than pretends.
Engineering
What it runs on
For the people who ask — and it is usually the first thing they ask.
Services
- ASP.NET Core on .NET 10
- Rust for the real-time tracking path
- PostgreSQL 18 with PostGIS
- Redis for rate limiting
Applications
- Next.js and React 19
- Ionic React for the mobile builds
- Android via Capacitor
- TypeScript throughout
Running it
- Docker on Oracle Cloud
- Caddy, with automatic TLS
- Cloudflare for DNS, CDN and storage
- One machine, sixteen containers
All of it — every service, every database — runs on a single machine behind a reverse proxy that manages its own certificates. That is a deliberate choice, not a limitation: one machine is one thing to understand, and the services were built so that moving any of them elsewhere is a change of one URL.
Work with us
Whether you need one of our products, a platform underneath something you are building, or a job.