Work

Real systems. Real businesses.

Most of this I've built inside a major European tech company — plus a real-estate copilot I've built end to end and a hackathon win. No invented logos, no borrowed metrics.

~0
People who rely on it daily
0
Fully-automated tracking pipelines
Daily
In production use across multiple markets
1st
Place at Fixathon, for a real-world tool
Featured case
Case · Enterprise document intelligence

A document-intelligence platform for a major European tech company.

A production AI system that monitors fast-moving external change in real time across the markets the business operates in — running automated trackers, enriching each item with an LLM over a hybrid SQL + vector retrieval layer, and delivering cited, self-updating briefings now relied on daily by a large in-house team.

The problem

A large team was tracking fast-changing information by hand across multiple markets — slow, easy to miss something that mattered, and a poor use of expert time.

What I built

A production pipeline that runs automated trackers across multiple markets, enriches each item with an LLM over a hybrid SQL + vector retrieval layer, and produces summarised, cited briefings for the team.

The result

Now in daily use across a large in-house team. Every briefing cites its source, so nothing has to be taken on trust — and nothing slips through.

Approach

RAGLLM enrichmentHybrid SQL + vectorAutomated monitoringCited output
How it works · the flagship pipelineLive
  1. Market sources
  2. Automated trackers
  3. Hybrid SQL + vector
  4. LLM enrichment
  5. Cited briefing

Documents and signals in → cited, self-updating briefings out. Illustrative of the production architecture.

More systems

Not a one-off. A pattern.

Real estate · Built

propdesk — from a pile of documents to a defensible bid

propdesk is the makelaar copilot I built for a small real-estate team. Drop in a property's documents and it reads them into a cited, verified record — catching what's wrong or missing, and cross-checking the m² against the official Dutch BAG register — then writes the NL/EN listing, answers buyers, and prices the bid against comparable sales with explained, comp-cited reasoning. The same retrieval and extraction patterns as my enterprise work, pointed at property. Built and verified on synthetic and real public data; not yet in production.

Tap to expand. Fictional sample — no real client data.
Enterprise · Internal tools

A single hub for a large in-house team

One place for everything a large, cross-market team actually needs: documents, resources, AI tools, knowledge and people mapping, news, and a way to ask a question and find the answer — or the right colleague — fast.

Also1st place at Fixathon — a build that helps people who'd just lost their jobs work out how to file claims and figure out their next step.View on LinkedIn ↗

The standard

Built for them. Now for you.

No invented logos, no borrowed metrics. When your project ships, it joins this page — with real numbers and your name on it, if you want it there.

Start a project →