Filip Van den Broeck
Filip Van den Broeck
Senior engineer and fractional tech lead based in Antwerp. 15 years building enterprise systems. The last two spent obsessing over what AI actually changes about software development.
The Story So Far
I spent 15 years deep inside enterprise engineering — most of it at BNP Paribas Fortis, where I worked as a senior developer and eventually as Technical Lead on Centric, a financial platform serving 100,000+ users. I also built DinnerGift from scratch: full-stack, end-to-end, production-grade.
I know what it means to ship software that cannot fail. I know what good architecture looks like under pressure. And I know how teams actually work — not how they're supposed to work on a slide deck.
The Shift
In the last two years, I've become obsessed with one question: what happens when AI becomes a first-class member of the development team?
Not AI as a fancy autocomplete. Not a chatbot bolted onto a product. AI as a genuine collaborator in the engineering process — one that handles research, drafts code, catches regressions, writes documentation, and surfaces insights while a human engineer drives the architecture and quality bar.
To explore this seriously, I built Hermes Agent — a local, autonomous AI coding assistant that runs entirely on my own infrastructure. Hermes operates in a persistent loop: it receives tasks, reasons through them, writes and executes code, reads test output, and self-corrects. It runs on open-weight models with no data leaving the machine.
The result? I can handle workloads that would normally require a team of two or three — with tighter feedback loops, better documentation, and more consistent code quality than pure manual work.
What This Means for You
I work with optimized AI workflows that let me move fast without sacrificing depth. Every engagement gets the full weight of 15 years of enterprise experience — not just on the code I write, but on what that code needs to become six months from now.
The pitch isn't speed. The pitch is speed and production-grade quality. Those two things usually trade off. AI-augmented senior engineering is how they don't.
Whether that means auditing your team's AI readiness, stepping in as a fractional tech lead, or standing up a self-hosted AI stack — the engagement is deliberate and focused.
If that sounds like what you need, let's talk.
What I Do Best
AI-Augmented Engineering
I build and use multi-agent AI workflows in real production contexts — not as experiments, but as a core part of how I ship software faster and with higher quality.
Quality Without Compromise
15 years of enterprise delivery means I hold a high bar. AI speeds things up — it doesn't lower standards. Every output is reviewed, tested, and owned.
Clear Communication
I translate between engineers, product owners, and stakeholders. No jargon walls, no hand-waving. Just honest, structured updates and decisions explained with context.
Full Ownership
I work independently and take responsibility for outcomes — not just outputs. If something breaks or needs rethinking, I'm already on it.
Systems Thinking
I look at architecture, team dynamics, tooling, and workflow as one interconnected system. Good engineering is never just about the code.
Good to Work With
Focused, reliable, and human. I bring energy to hard problems and stay level-headed when things get complicated.
Experience Highlights
Technical Lead — BNP Paribas Fortis / Centric
Led architecture and delivery for Centric, a financial platform used by 100,000+ users. Owned backend and frontend across Java, Spring, and Angular. Managed team quality practices and agile delivery.
Owner & Full-Stack Developer — DinnerGift
Built end-to-end from concept to production: React/Next.js frontend, Express REST API, MongoDB, and CircleCI CI/CD pipeline. Ran independently for years as a commercial platform.
Builder — Hermes Agent & Multi-AI Workflows
Designed and built a locally-running autonomous AI coding agent. Handles task decomposition, code generation, test execution, and self-correction — all on self-hosted infrastructure with open-weight models.