01
I’m a full-stack developer in Liège who likes the seams of a product: where raw data becomes a dependable service, and that service becomes an interface someone can understand.
My recent work ranges from BizComp — a business intelligence platform built on official Belgian data — to public-facing tools for energy auditing and medical scheduling. I care about resilient backends, accessible interfaces, and systems that remain understandable after launch.
02
Full-stack web platforms, backend architectures, data pipelines, and AI integrations.
Featured system
Business intelligence platform built on official Belgian enterprise data. Ingested the full KBO registry, built a multilingual AI conversational agent (FR/EN/NL) for financial/company analysis, and automated balance sheet extraction.
Full-featured web application for medical appointment booking. Patients filter doctors by specialty and book interactive time slots, while practitioners manage their schedules via a dedicated dashboard.
03
A closer look at the engineering decisions behind BizComp Intelligence. Open a note for the implementation detail.
12intent types
User queries are often ambiguous: 'Colruyt finances' could mean a quick health check or a deep ratio analysis. I built a classifier mapping requests to one of 12 structured intent types with conversational context normalization. The entity resolver queries the Crossroads Bank for Enterprises in parallel with classification to minimize response latency.
5pipeline stages
Each request flows through more than 5 asynchronous stages (classify, resolve, plan, execute tools, compose). I implemented request budgets with abort propagation: if the client disconnects or times out, in-flight LLM calls and SQL queries are cancelled immediately, preserving backend server resources.
3retries
When multiple rapid or concurrent requests hit the same session, state corruption is a significant risk. I implemented optimistic concurrency control with a version column on PostgreSQL. In the event of a lock conflict, the system reloads the newest session state and retries up to 3 times automatically.
To prevent hitting provider inference quotas under heavy load, I created a shared LLM queue manager with token bucket rate limiting and intelligent load shedding. The system prioritizes interactive user sessions and gracefully sheds low-priority traffic before downstream saturation occurs.
04
Have a question about a project, an opportunity, or just want to connect? Feel free to send a message.
