Geodata Engineering &
AI/ML Automation
for ops-heavy teams.
I design and ship production data systems and AI automations — pipelines, models and internal tools that replace manual work. Currently working across geospatial analytics, insurance, logistics, public sector and real estate.
Core competencies.
Specialized skills for production data systems, AI automations and geospatial pipelines — built to ship and stay shipped.
Machine
Learning & AI
Production ML and AI — predictive models for valuation, rent and flood risk (90%+ accuracy), plus LLM pipelines and tool-using AI agents that automate research, classification and reporting.
Data
Engineering
Robust data pipelines for European REITs, funds and institutional investors — geodata, alternative and unstructured inputs running on a distributed, dockerized Airflow stack.
Geospatial
Analysis
Advanced spatial analytics — agent-based gentrification modelling, street-level computer vision, and spatially-aware data science (downsampling, significance testing) that yields statistically robust results.
Internal
Processes
The backend of a business — sales systems, CRMs and AI-powered automations that replace manual work and scale with the team.
Where I tend to work.
Domains I've spent time in — academically, professionally or both. Each has its own data quirks and its own definition of "good enough."
Geospatial Analytics
Spatial pipelines, H3 indexing, parcel-level models, alt-data ingestion.
Insurance & Climate Risk
Exposure modelling, flood and CV pipelines, claims triage automations.
Logistics & Mobility
Routing, depot siting, last-mile, ops dashboards that replace ten manual reports.
Public Sector & Planning
Mobility, climate adaptation, infrastructure resilience and parcel-level analysis.
Real Estate & Investment
Valuation, gentrification ABMs, alt-data for REITs, funds and asset managers.
Selected work.
Things I've built or studied — at university, in research, and in my own time. Not a portfolio of client deliverables; a portfolio of thinking.
Gentrification agent-based model
Ten-year simulation of neighbourhood turnover in European cities, calibrated on Kadaster and CBS microdata.
Parcel-level flood-risk classifier
Supervised classification on LiDAR and rainfall-radar features at parcel level — written up as the bachelor's thesis.
Hedonic rent model · 15 EU metros
Parcel-level hedonic rent benchmarks built on PostGIS and gradient boosting across European metros.
LanguageBuddy
A small conversational language-practice app: LLM-driven correction, spaced repetition, and a habit-friendly UI.
Façade & commerce signals from street-view
CNN extraction of façade and commerce features from ~4M street-view frames across six EU cities.
Recent thoughts.
Essays, tutorials and field notes. View all →
Selected papers.
Recent output on real-estate econometrics and geospatial methods. View All →