Seb Sebahattin K.

I build the spatial data layer and the map that sits on top of it. Most of my work is PostGIS and dbt pipelines that turn messy geodata into something a product can query in milliseconds, plus the interactive maps that make it usable. 🗄️ SPATIAL DATA ENGINEERING PostGIS modeling, spatial ETL, dbt warehouses. On my largest ongoing project: • 377-model dbt project on PostGIS, 8.6 million parcel records, statewide • Index strategy, incremental builds, Airflow orchestration • Rewrote a join from 92M nested-loop comparisons to an indexed unnest join, measured 126x faster • Cold vector tiles taken from 15-25 seconds down to 0.3-0.7 seconds 🗺️ INTERACTIVE MAPS • Mapbox GL JS, Leaflet, MapLibre, OpenLayers, deck.gl • Clustering that stays fast at scale, in-database vector tiles • Choropleths, heatmaps, 3D terrain, custom canvas layers when the standard ones are too slow ⚙️ FULL-STACK GIS PROTOTYPES When you are not sure yet if something is buildable, I build the small version first and measure it. • Python (FastAPI, Django/GeoDjango, shapely, GDAL) or Node.js on the backend • PostGIS underneath, a real map on the front 🛰️ SATELLITE & REMOTE SENSING • Google Earth Engine, Open Data Cube, Sentinel-1 SAR, Sentinel-2, Landsat, Himawari SST • 36-year Landsat surface water history for Australian Ramsar wetlands • Sentinel-1 radar screening for oil spills at sea 🧰 STACK • Data: PostGIS, PostgreSQL, dbt, Python, Node.js, Airflow, GDAL/ogr2ogr, spatial SQL optimization • Maps: Mapbox GL JS, Leaflet, MapLibre, OpenLayers, deck.gl, vector tiles (MVT), PMTiles, GeoJSON • Backend: Django/GeoDjango, FastAPI, Express, REST APIs • Infra: AWS, Azure, Docker, CI/CD, GeoServer, QGIS • Raster: Sentinel, Landsat, GeoTIFF, raster processing, zonal statistics 🤝 HOW I WORK I am not only a developer. I tell clients what will not work before they spend money on it, and I would rather measure something than promise a number I cannot back. Several of my clients have stayed with me for years, one of them for over 900 hours on a single ongoing project, and that is the part of my record I am most proud of. On Upwork: Top Rated Plus, 100% job success, 78 projects, 2,284 hours. Live demos of my mapping work are in the portfolio section below. If you have spatial data that is slow, messy, or stuck in files nobody can query, or you need a map that actually performs, send me the details and I will tell you honestly whether I am the right person for it.