Projects & papers.

Eight pieces of work across the four lanes: production systems, shipped products, and the research they make possible.

conversationlemmatizeSM-21d6d15dnext-day6,200 vocab · A1–C1
§ 04.01AI productPROJECT

LanguageBuddy: AI language tutor

A self-hosted AI language tutor for Dutch, Italian and Spanish, built on language learning research that includes a chat or voice-call LLM tutor and SM-2 repetition. Every mistake is captured into a spaced-repetition queue that drives the next day's exercises and uses a real-news scraper for reading learning. Follows the CEFR framework with 6,200+ vocabulary entries and helps users reach CEFR alignment faster.

FastAPILLMTTSSQLiteDocker
t=0t=10yattractiveness ↑affordability ↓parcels: AMS · UTRECHT · MILAN
§ 04.02MSc Thesis · BocconiRESEARCH

Gentrification agent-based model

My MSc Thesis research on gentrification, where I built an agent-based model of neighbourhood change: households interact with neighborhoods in the treated cities Amsterdam, Utrecht and Milan based on their own income, the affordability of the neighborhood and the attractiveness, based on elements such as streetview imagery 'beauty' classification, GTFS connectivity, greenery and neighborhood sentiment online. A two-tenure social-housing extension is in progress.

Python/GeoPandasLarge Vision ModelPostGISAgent-based Modeling
DAGingest726 GBguardtransformqueuecelery workers
§ 04.03EngineeringPROJECT

Research pipelines as production systems

Every personal research project I undertake uses Airflow 3 data pipelines that are hosted in this project. It can scale from a laptop to a multi-machine CeleryExecutor cluster. This project was built with idempotency guards, custom operators and agentic monitoring.

AirflowCeleryDockerCIDuckDB
p(sponsor)86% F10.54:326:05sentence-T5 → BiLSTM · 38,600 videos
§ 04.04NLP · sequence taggingPROJECT

SponsoredBye: sponsor-segment detection

A text-only sponsor-skipper for YouTube, built before YouTube Premium shipped one: sentence-T5 embeddings feed a BiLSTM sequence tagger that flags sponsored sentences and maps them back to timestamps, cutting segmentation error from 99% to 16% WindowDiff.

TensorFlowBiLSTMsentence-T5MongoDBHuggingface
SAM224pxResNet501/638.8 MB on-device
§ 04.05Mobile MLPROJECT

FishFinder: photo-to-species ID

A Flutter app that identifies 63 Dutch fish species from a photo, fully on-device, and fills a Pokédex-style FishDex as you catch them. The training pipeline: ~3,000 hand-annotated photos masked with Segment Anything Model, then a fine-tuned ResNet50 compressed to an 8.8 MB TFLite model for on-device use.

Flutter/DartTFLiteResNet50Segment Anything Model (SAM)Firebase
RF 97.5%33 featuresd(river)Δhimperv. 500m–5kmw.l.
§ 04.06BSc Thesis · UvARESEARCH

Predicting flooding risk from local features

My BSc thesis, looking at whether flood risk can be explained by local features instead of a black-box hydrodynamic simulation. Collected 33 features across ~45,000 European locations, such as ground imperviousness, ground type and distance to river, across 100GB+ of data and ran a Random Forest to find their relation. Findings: 97.5% on the binary 20-year flood question, with surrounding imperviousness and relative height doing most of the work.

scikit-learnRandom Forestraster dataGIS
945 min15 min38 markets · EU/NA/APACparcel score 0–100
§ 04.07Developed at KR&AKR&A

Connectivity & walkability scoring

A project led and conducted while working at KR&A: a saturation-validated connectivity and walkability score at parcel resolution, rolled out across 38 EU/NA/APAC markets, working with TBs of data and 100s of data sources.

PostGISS3Dijkstra & Network ScienceDistributed Compute
monthlyEurostat Qbase=10013 countries
§ 04.08Developed at KR&AKR&A

Monthly house-price index · 13 EU countries

A project conducted while working at KR&A: a web-scraping pipeline across 13 countries feeding a log-price hedonic regression; monthly indices tested as disaggregation indicators for Eurostat quarterly HPIs.

PythonScrapy/SeleniumPostGISMongoDBR

Papers.

Working papers, preprints and method write-ups on research in urban dynamics, network analysis, geospatial methods and data science.

01
2025-03-05
US vs EU: does training-data geography matter for autonomous-driving object detection?A controlled 2×3 fine-tuning study on YOLOv3/YOLOv8: US dashcam data transfers roughly nothing to European streets (+0.001 mAP vs +0.153 for in-domain data), and bicycle detection collapses without EU ground truth.Download PDF ↓
WORKING-PAPER
02
2026-07-09
When Does Metro Infrastructure Capitalize into Property Prices? Phase-Decomposed Difference-in-Differences Evidence from Seven European CitiesWhen do new metro lines show up in house prices? Across 42,000 observations from seven European cities, the market pays on delivery, not on promises, and the number worth defending is Milan's +167 EUR/m² within Ring D.Download PDF ↓
IN PROGRESS
03
2026-08-14
Calibrating Free Postcode Boundaries from OpenStreetMap: A Transferable Seed-Density Accuracy CurveFree postcode boundaries built from OpenStreetMap addresses, with a calibrated accuracy curve that predicts how good they are before any reference map exists. Tested across five countries, deployed on all of Italy.Download PDF ↓
WORKING-PAPER
04
Social housing impact on the gentrification ABM
IN PROGRESS
05
Urban heat island research
IN PROGRESS
Get in touch
Projects & Papers | Ian Ronk