About Me
Hi there! I'm a skilled freelance AI Engineer and Data Scientist with a passion for building AI systems that actually work in production, not just impressive demos. With a PhD background in AI and astrophysics, I bring both deep theoretical knowledge and a pragmatic engineering mindset to every project.
My recent work includes building RAG-powered Q&A systems for enterprise clients, developing secure chatbots on technical documentation, and creating satellite imagery segmentation models for vegetation monitoring. I specialise in LLMs, agentic systems, and end-to-end ML pipelines that deliver real business value.
I have a strong foundation in Python, PyTorch, and production AI tooling including vLLM, Huggingface Transformers, RAG pipelines, and agentic frameworks. I deploy on Docker, Kubernetes, and Azure, and I'm an active user of coding agents like Claude Code and Codex. I also love experimenting with local LLMs on my Mac.
If you're looking for someone who can take AI from prototype to production and cares about building systems that genuinely help people, let's connect!
Resume
Experience
AI Engineer / Data Scientist / ML Engineer
- • Self-employed, building AI and ML solutions
Senior Data Scientist
- • Developing and productionizing real-world AI models
- • Scaling up AI agents securely for enterprise use
- • Advising on AI strategy, architecture, and responsible deployment
Medior Data Scientist
- • Built LLM-based Q&A system for service desk employees
- • Developed secure RAG chatbot on technical documentation
- • Created vegetation monitoring system using satellite AI
PhD Researcher, AI in Astronomy
- • Developed DeepGlow: ML emulator achieving 10,000× speedup
- • Built production-grade neural network using TensorFlow/Keras
Astronomy Teacher
Education
MSc High-energy Astrophysics
- • Thesis: Simulating the gravitational-wave memory effect (cum laude)
FNWI Honours Academy
- • Top five percent of students accepted
BSc Physics and Astronomy
- • Thesis: Inferring black hole mass distribution from gravitational-wave detections (cum laude)
Projects
LLM Q&A System for Service Desk
Built a retrieval-augmented Q&A system to support service desk employees with mild intellectual disabilities. Implemented multi-stage retrieval pipeline with query rewriting, semantic search using Milvus, and custom RAG pipeline. Self-hosted all models on Nvidia DGX using Docker and vLLM.
Secure RAG Chatbot on Technical Documentation
Developed MVP chatbot for querying technical documentation across PDFs, Word, and Excel. Built universal parsing pipeline with marker-pdf, implemented hybrid search (semantic + BM-25), and deployed on-premise with Streamlit, Milvus, and PostgreSQL. Later scaled to Azure Kubernetes Service.
Vegetation Monitoring with Satellite AI
Led team to productionize transformer-based segmentation model for monitoring vegetation in Rhine and Maas floodplains. Developed MLOps pipelines with Airflow and ArgoCD on Kubernetes, processing ~1000 km² of satellite images monthly. Achieved 80% accuracy for legal compliance reporting.
End-to-End AI Pipeline for Business Discovery
Designed LLM-driven automation pipeline for SPAIK that transforms meeting transcripts into structured project proposals and functional React prototypes. Used Gemini Pro 2.5 for document processing and built custom tooling for AI-generated code rendering. Reduced multi-day workflow to minutes.
DeepGlow ML Emulator
PhD project developing production-grade neural network emulator for astronomical simulations using TensorFlow/Keras. Achieved ~10,000× speedup with millisecond inference times. Implemented custom loss functions, cyclic learning-rate scheduling, and released as open-source Python package.
Eredivisie Predictions Website
Full-stack web application for football predictions competition with user authentication, real-time score updates via third-party API, and automated points calculations. Built with SvelteKit, SQLite/Drizzle ORM, and Tailwind CSS. Self-hosted on cloud VPS with Coolify for CD.