Frontline commander at Brandweer Zone Antwerpen (Antwerp Fire Department) · BIRD team (Drones, IT, Robotics & Digital Technologies) · Building applied AI systems
I work where emergency response, technology and implementation meet. For over 16 years I have served with the Antwerp Fire Department, currently as a frontline commander and operational leader of the BIRD team. There I lead teams during real incidents and drive the practical adoption of drones, robotics, data and digital technology in public safety.
In 2026 I completed the Ironhack AI Engineering bootcamp, covering machine learning, computer vision, deep neural networks, large language models and AI-driven automation. The projects below apply that knowledge to real operational problems. AAR and IncidentIQ are built for my own field, public safety. The other projects are practice builds: each one exists to learn and drill a specific technique, such as voice agents, document search (RAG), agent workflows and model fine-tuning. All of them are built to be reliable, traceable and usable by the people who depend on them.
My approach: start from the operational need, test what is possible under realistic conditions, and turn technology into something people can actually use in practice.
Operational problems I ran into in emergency response, tackled with the techniques I built up in the skill-building projects below.
| Project | Description | Stack |
|---|---|---|
| AAR | AI-assisted After Action Review. Turns everyone's reports into one overview the facilitator uses to lead the debrief: summary, timeline, what went well, points for improvement and lessons learned, each linked to its source sentence. Built for fire services, adaptable to any team that runs debriefs. | Claude · FastAPI · SQLAlchemy · Pydantic |
| IncidentIQ | Turns incident training videos into searchable knowledge, debrief documents and scenarios for XVR, a VR training simulator. | LangGraph · RAG · Pinecone · LangSmith |
Built to learn and practise specific techniques. Each one is a complete, working project, but the goal is the technique, not the product.
| Project | Description | Technique practised | Stack |
|---|---|---|---|
| Cordon 🆕 Latest project · October 2026 |
AI gateway between a company's apps and models like ChatGPT: masks personal data before it leaves, blocks prompt injection, keeps each team within its budget and sends a request to a cheaper model only where that model was proven as good - 80% saved on a simulated working day, measured live. | LLM cost optimisation and routing, guardrails, evaluation with tests fixed in advance | FastAPI · LiteLLM · scikit-learn · Docker |
| Lisa | AI voice receptionist that answers calls, books appointments and escalates to a human. | Real-time voice agents, tool calling | OpenAI Realtime · Twilio · FastAPI · PostgreSQL |
| Servix | Field-service assistant that troubleshoots HVAC fault codes from service manuals, always citing the source page. | RAG with source citations, vision, guardrails | GPT-4o Vision · Qdrant · RAG · Docker |
| Lead Scraper | B2B prospecting pipeline: finds, scores and drafts outreach for human review. | Agent workflows, human-in-the-loop | LangGraph · Claude · Playwright |
| Amazon Review Analyzer | NLP pipeline on 46,000 product reviews: fine-tuned a BERT sentiment classifier (77.0% → 79.4% accuracy), clustered products with sentence embeddings and generated buyer's guides with a local LLM. | Fine-tuning, embeddings, local LLMs | Hugging Face Transformers · PyTorch · Llama 3.2 · Streamlit |
Bootcamp exercises: ironhack-ai-engineering-bootcamp
- Operational leadership: station commander and incident leader, coordinating multidisciplinary teams under pressure
- Public-safety innovation: testing drones, robotics and digital systems in realistic conditions and turning lessons from exercises and incidents into procedures and training
- Training & knowledge sharing: Incident Command System (ICS), and firefighting strategy for urban, tunnel and heavy-industrial environments
- Belgian Defence: experience in a structured military environment
- Ironhack AI Engineering bootcamp (2026): 3-month full-time programme covering machine learning, computer vision, deep learning, LLMs, RAG and AI agents, with hands-on projects every week
- AI & LLMs: OpenAI (Realtime, Vision), Anthropic Claude, LangChain, LangGraph, Hugging Face, RAG, agents & tool calling, evaluation (RAGAS, LangSmith)
- Machine learning: model fine-tuning (BERT; LoRA / QLoRA with PEFT in coursework), scikit-learn, PyTorch, Keras, NLP, computer vision (coursework)
- Engineering: Python, FastAPI, PostgreSQL, Qdrant, Pinecone, Docker, pytest, Pydantic
A live demo, a project presentation or answers to any questions about these projects are available on request.