30x faster zonal stats with cloud-native COGs and exactextract
Most working GIS code runs about 30 times slower than it has to.
Al Hotimski
Geospatial Data Scientist at The Conservation Fund with an MS in Remote Sensing and Geospatial Sciences (Environmental Data Science). I build data systems and analytical models that make sense of the landscape: land ownership, urban growth, fire risk, and conservation opportunity. The tools I have shipped at TCF have cut analysis time from 20+ hours to 10 minutes and reduced manual QA workloads by 93%.
What I Do
Geospatial Analysis
Land cover classification, change detection, and spatial data modeling. ArcGIS Pro, QGIS, Google Earth Engine, geopandas, rasterio. Recent: 17 years of Landsat across LA County's fire hazard zones.
Machine Learning
Supervised classification, time-series change detection (CCDC), and fuzzy entity resolution applied to spatial and environmental data. scikit-learn, pandas, geopandas.
Conservation Tech
End-to-end pipelines for land trusts and conservation orgs. USDA NRCS ACEP grant screening that replaces manual Web Soil Survey queries with one regional run. LLC-resolved parcel ownership and boundary change tracking across Southeast U.S. timberland.
Projects
cogsieve
An open-source Python library that filters polygons by fractional class coverage of categorical rasters, reading windowed pixels directly from remote Cloud-Optimized GeoTIFFs instead of downloading scenes. Three runnable demos on public data: solar siting, tree equity, wildfire risk.
Open source (MIT), CI-tested, 30x faster than the standard pure-Python zonal-stats workflow. Live demo · GitHub
Python rasterio geopandas exactextract COG Streamlit
parcel-lineage
An open-source library that reconciles messy county parcel owner names into their true holdings and tracks ownership change between snapshots. Runs live against public ArcGIS parcel services.
Open source (MIT), typed and tested. On real Adirondack data, it collapses 711 raw owner strings to surface the largest true timberland holders. GitHub
Python rapidfuzz pandas ArcGIS REST Entity Resolution
Land Ownership and Boundary Change Detection
An automated pipeline that resolves LLC ownership structures in county parcel records and detects ownership and boundary changes between time-stamped snapshots using spatial overlay analysis and symmetric difference.
93% reduction in manual QA review. Longitudinal ownership intelligence across the Southeast U.S.
Python SQL Data Harmonization Relational DB Spatial Overlay ReGrid
ACEP Easement Eligibility Screening Tool
Automated USDA NRCS ACEP easement grant screening by applying SSURGO and NLCD raster criteria via REST services.
Analysis time: 20+ hours → 10 minutes. Projected $12,000/yr savings.
Python ArcGIS Pro REST APIs SSURGO NLCD
Urban Growth in Fire Hazard Zones
A 9-class Landsat change detection (2007-2024) paired with a full Olofsson accuracy assessment. New development expanded into LA County’s very high fire hazard zones before the hazard designation caught up; the mapped extent corrects from ~16,600 ha to ~5,700 ha once bias-adjusted.
GEE Random Forest Landsat Accuracy Assessment
Skills
| Domain | Tools |
|---|---|
| Languages | Python, R, SQL, JavaScript, Bash |
| ML & Analysis | scikit-learn, supervised classification, fuzzy matching, data harmonization, entity resolution, time-series, change detection (CCDC) |
| Data Engineering | pandas, NumPy, relational databases, ETL pipelines, Git |
| GIS & Remote Sensing | ArcGIS Pro, QGIS, Google Earth Engine (cloud-native), geopandas, shapely, rasterio, Landsat |
| Visualization | Power BI, ArcGIS Online, matplotlib |
Get in touch
Open to geospatial data science, remote sensing, and spatial ML roles.