About

AI made code cheap. It didn't make software good.

I'm a software engineer. I take projects from the first conversation to production: web apps, mobile apps and business systems, with every part tested before it goes live. AI made software fast and cheap to build. It did not make it reliable. That gap is the work I do: you get the software, and the evidence that it holds up.

The project that taught me to own a whole system was a restaurant platform in Belgium. Six connected applications, built on my own and still in production: an ordering app for customers, dashboards for the staff, an app on the owner's phone, and a printer in the kitchen that prints every order without anyone touching it. On a build like that there is nobody to hand a layer to, so every failure is mine to find.

Today coding agents write most of my code. They work against a specification and an architecture I write first, so they implement the design rather than decide it. Tests come from the API contract, not from the code an agent wrote. A check passes or fails on its result, not on what anyone reports about it. And every mistake I catch becomes a permanent check, so it only costs once. The result is AI speed with a codebase a person can still own.

Process

How I work

Understand

I find out what the software must do before anything is built, and write it down so we agree on it.

Build with AI

AI tools write much of the code, inside rules and patterns I set. That's where the speed comes from.

Check everything

Every part is tested against what you asked for. Nothing ships because the AI said it works.

Own it

I deploy it, run it and answer for it. You get the evidence and the full record, and you own both.

Perspective

What I believe

Polish is not proof.

A demo that looks good is not evidence. Evidence is software that was run, tested and checked against what you needed.

AI can do the work, but a person must own the outcome.

An agent can type the code. It can't answer for it. I decide in advance which calls stay with me: money, anything that can't be undone, and anything that touches your users' trust.

An AI agent should never have more power than the person it acts for.

When software acts for a user, it acts with that user's permissions and nothing more. If the user can't touch a record, neither can the agent.

Working style

How I prefer to work

  • Remote-first. I do my best work with autonomy and deep focus
  • Asynchronous communication. Fewer meetings, more building
  • Full ownership. I take a problem and come back with a working product
  • Contract-based or flexible engagements, with aligned incentives and a clear scope
  • Results over hours. I care about what gets shipped, not what time I logged in

Technology

Tech I work with

AI and verification
Claude CodeAgentic coding workflowsSpec-first promptingContract testsAutomated quality gates
Frontend
Next.jsReactTypeScriptTailwind CSSshadcn/uiTanStack QueryTanStack TableReact Hook FormZod
Backend
Node.jsExpressREST APIsJWT AuthenticationRedisSocket.IOWebSocketsZod ValidationSwagger / OpenAPIGoPrismaBullMQVitest
Mobile
React NativeFirebase Cloud Messaging
Desktop
Electron
Databases
MongoDBPostgreSQL
Infrastructure
Linux VPSNginxPM2CloudinaryGitGitHub ActionsDockerPrometheusGrafana