Four roles, one stack.
Most GIS work stalls at a handoff: data to services, services to apps, apps to models. I work every layer, so nothing waits on someone else.
Web GIS apps people actually use, under pressure.
Custom web maps, Experience Builder apps, dashboards and the APIs behind them, built for field crews, operations teams and leadership.
- Spatial web portals used in real time during Hurricane Milton for impacts, outages and crew deployment.
- Storm-pilot app tracking UAV pilots live, plus an editor that let non-GIS staff update outreach sites themselves.
- Freelance: an offline-first field data collection app for a forestry client, and a 3M-address geocoder that replaced a paid service.
Enterprise GIS that stays up, and stays clean.
ArcGIS Enterprise and Portal administration, feature services, multi-user editing and the PostgreSQL / PostGIS databases underneath.
- Published a taxonomy as a feature service with editor tracking and concurrent multi-user editing, plus a read-only REST API for vendors.
- Moved 8M+ images and 65M+ annotations into an FME-to-PostGIS pipeline with versioning, cutting 100+ columns to ~50.
- Ran PostgreSQL / PostGIS databases, Cloud-Optimized GeoTIFFs and automated QA in FME, ArcPy and ModelBuilder.
Tools other GIS people pip install.
Published Python packages, an MCP server for AI agents, and core contributions to a flood modeling framework. Built in the open, documented, tested.
- road-centerline on PyPI: CRS-aware road centerline extraction from polygons, worldwide.
- geomlint on PyPI: CI-first validity, topology and CRS checks for vector data.
- One of four core contributors to PyWMP, an open-source flood modeling framework, and author of HazardMCP.
Models that leave the notebook and go to work.
Computer vision for infrastructure, risk models and flood research, evaluated properly and deployed to production.
- Trained and shipped a switch-state classifier to production on AWS SageMaker for a utility inspection program.
- Benchmarked YOLO, RT-DETR, Detectron2 and more on precision, recall and F1; standardized the approach for vendors.
- Peer-reviewed flood hazard and social vulnerability research (ISPRS IJGI, 2024) and FEMA-funded H&H assessments.
Open source, shipped and live.
Built on my own time and free to try. Most run in the browser with no API key.
Eight bands, one composite.
Like a multiband raster: each band holds one kind of skill. Pick a band to read it, or a role to see its composite.
Where I've built it.
Florida Atlantic University · 4.0 GPA
BSc Geomatics Engineering, Tribhuvan University
Dhungana & Liu, ISPRS IJGI 2024 · DOI ↗
PyWMP core contributor · docs ↗
NASA ARSET (12) · Esri (8) · DeepLearning.AI Machine Learning and Deep Learning specializations