Everyone has the same AI models — the work is the system around them. I build that system on-site: the tools, the guardrails and the handover that turn AI from an impressive demo into something your team actually uses. The projects in my portfolio are exactly this work, already shipped.
Pick the right tools. I choose what actually solves your problem, not what wins the demo — tested against your real work before anything goes near production. The same judgement behind the on-prem AI server and the AI Monitor in my portfolio.
Wire them in safely. AI plugged into your existing setup without breaking it: clear limits on what data goes where, a human check wherever the risk is real, and GDPR, the AI Act and AML considered from the start, not bolted on after.
Hand it over. Your team ends up owning the system — every build ends in a clear yes/no and a plain-language handover. No lock-in, no permanent dependence on me.
Finding the real root cause, not the symptom you were told about. I talk to people at every level and map how the work actually flows — versus how it's supposed to.
Designing an AI setup that fixes what's actually broken — the right tools, who owns what, and a plan sized to what your team can really take on.
Running the build with zero tolerance for scope creep — managing vendors and agencies, onboarding the team, and keeping progress on the plan, not the demo.
Handing the finished system to your own people once it holds — with the documentation to run it. No lock-in; staying longer than needed is a failure, not a business model.
Built, not promised
Real systems, running today — for clients and for my own practice. Not slideware.
Where does your data actually go when you hand it to an AI? An interactive site that maps every major AI platform and gives each one a clear GDPR & AI Act verdict — built once, kept current, in English and French. The answer to "is it safe to use this?", on a page you can hand to a client.
Buried numbers become live dashboards — spend, KPIs and status as colour-coded gauges that refresh themselves from your inbox and tools. No more month-end spreadsheet archaeology.
Built for a fiduciary →
A multi-persona agent dashboard that compiles a daily intelligence brief — several specialist AI personas, one decision-ready digest each morning. A proof-of-concept for agentic, always-on reporting.
Many personas · one brief →
AI that never leaves the building. A local model server for regulated, can't-touch-the-cloud work — zero data egress, no third party, full audit trail. For when "just use ChatGPT" isn't allowed.
Zero data egress →
One meeting recording becomes a costed, structured 20-page proposal — in minutes instead of days. The conversation goes in; a decision-ready plan comes out.
Minutes, not days →
An AI engine that finds, verifies and scores real prospects against your ideal profile — no hallucinated leads. Delivered as a setup mission you keep and re-run in-house.
Search · verify · score →
A team of specialist agents — research, numbers, build, compliance, and a devil's advocate — that turn any decision into a sourced brief ending in a clear go / refine / drop. Multi-agent, orchestrated, not a single chatbot.
Decision → sourced brief →
Paste a role description; get a CV and cover letter tailored to it — a web front end on a hosted backend. The same pattern that tailors any document to any target: proposals, briefs, applications.
Source in · tailored doc out →
A chat that recommends films and shows from your own viewing history — GPT reasoning over the TMDb catalogue. A clean demonstration of the pattern that matters in business: AI sitting on top of your data and being genuinely useful.
Your data · useful answers →
A mobile-first agent that scans Belgian ticketing sites every week and surfaces concerts for the artists and venues you care about. The same watch-and-alert pattern that tracks prices, filings or competitors for a business.
Watch · detect · alert →I read every message myself. Response within 2 business days. You bring the scope, I bring the plan.
Initiate Contact