Software architect AI systems Roosendaal, NL

Systems that hold up.
Models that behave.

I design backend, mobile and API systems for enterprises, and build the tools and context that make LLM agents reliable in production. When an agent fails, I fix the system around the model first.

Alex Luncan
Focus
Backend, API, mobile
AI
LLM agents, tooling, evals
Shipped
Enterprise production tools
01 — Principles

How I build

01

Boring where it can be, novel where it must.

Proven infrastructure underneath, invention only where it changes the outcome. Novelty is a cost to be spent deliberately.

02

Contracts outlive the code inside them.

APIs, schemas and boundaries between systems are the durable design. Frameworks are replaceable; interfaces are not.

03

If you cannot inspect it, you cannot fix it.

Traces, structured logs and evals are part of the product. This is doubly true for anything with a model in the loop.

04

Fix the system around the model first.

Context, tools and orchestration are cheap to change and easy to test. Fine-tuning is neither. Start where the leverage is.

02 — Agents

Fix the system around the model before the model.

Most agent failures are not model failures. They come from what the model was shown and what it was allowed to do. Pick a failure to see where it originates in the loop and the cheapest layer that fixes it.

Observed failure
The agent loop
01
Context
System prompt, retrieved documents, tool results, conversation state. Everything the model can see.
Fix here
02
Model
The foundation model. Reasons only over what is in the window and the tools it is offered.
03
Tools
Schemas, descriptions, validation and the errors returned. Defines what the model can do.
04
Environment
Orchestration, step budgets, checkpoints, stop conditions. Controls the loop itself.
Result feeds back into context Loop repeats until done
Diagnosis

Answers from stale or missing facts

The fact the model needed was never in its window. It filled the gap the only way it can: by guessing.

Fix

Retrieve and inject the data. Add freshness metadata to results and tell the model what it does not know. No weights involved.

Escalation ladder — cheapest first
01 Context Minutes
02 Tools Hours
03 Orchestration Days
04 Fine-tune Weeks

Fine-tuning is the last rung. It is slow, it freezes behaviour into weights, and it is only justified when the failure is stable, measured, and out of reach of the three layers above it.

03 — Stack

What I work with

Languages
  • C#
  • PHP
  • Kotlin
  • JavaScript
  • Java
Frameworks
  • .NET
  • Laravel
  • Next.js
  • Node.js
  • Unity
  • Tailwind CSS
Data
  • MSSQL
  • MySQL
  • PostgreSQL
  • MongoDB
  • Firebase
Platforms
  • Azure
  • Linux
  • Windows
  • Android
  • Active Directory
AI systems
  • LLM APIs
  • Tool calling / MCP
  • Retrieval (RAG)
  • Evals & tracing
  • Agent orchestration

Have a system that needs to hold up?