2026-09-02 –, Ran1
verture Maps Places offers 53 million POIs with a clean taxonomy — but what can you actually do with them? In 50 lines of Python and DuckDB, I turn raw Overture places into a visual neighborhood typology. Before/after, pipeline, and a ready-to-run notebook.
This is a one-idea, four-minute talk.
The idea: every neighborhood has a hidden fingerprint in its POI mix. Overture Maps Places — the new open dataset combining Meta, Microsoft, Foursquare, and OSM sources — gives us 53 million points with a standardized category taxonomy. That makes neighborhood profiling dramatically simpler than raw OSM tagging.
I will show one before/after image: thousands of raw Overture points versus the same city colored by discovered neighborhood types.
Then one workflow slide: DuckDB spatial query → H3 hexagons → POI feature vectors → UMAP + HDBSCAN → typology map.
Then one takeaway: a public Jupyter notebook. Replace the city name, run, done.
Stack: Python, DuckDB, GeoPandas, hdbscan, umap-learn, H3.
Overture Maps Places schema: https://docs.overturemaps.org/guides/places/
HDBSCAN docs: https://hdbscan.readthedocs.io/
UMAP docs: https://umap-learn.readthedocs.io/
H3 spatial indexing: https://h3geo.org/
Python, DuckDB, GeoPandas, hdbscan, umap-learn, H3, Overture Maps, Jupyter
I make my conference contribution available under the CC BY 4.0 license. The conference contribution comprises the abstract, the text contribution for the conference proceedings, the presentation materials as well as the video recording and live transmission of the presentation:
Senior Geospatial Analyst at Rockup, building neighborhood analytics tools. Previously Geospatial Researcher (R&D) at Habidatum, developing cross-country urban mobility pipelines for European policy institutions — OD matrices, temporal land-use profiling, service accessibility mapping across 16 countries. Former geospatial data scientist at Yandex, where I built GeoAI prediction models and led spatial feature engineering for service expansion. Invited lecturer on geospatial data science (MIPT Deep Learning School) and QGIS (RheinMain University, Germany). Jury member at IAAC Barcelona. Daily tools: Python, GeoPandas, PostGIS, QGIS. Admitted to MSc Geomatics at TU Delft. I run URBAN_MASH, a geoanalytics community (2,200+ subscribers).