Deployed and maintained 3 geospatial web applications, shipping 2 production cycles per week while incorporating researcher feedback.

Independently gathered requirements from non-technical researchers, translating ambiguous goals into defined specs.

Built reusable, modular UI components adopted across 3 applications, reducing code duplication.

Interactive GIS application simulating radar placement and quantifying coverage improvements across USGS basins.

Engineered backend raster processing across 47M+ grid cells per request with sub-4 second response times.

Client-side MRMS data viewer accessing 60+ NOAA weather products from S3, rendering animated overlays on Google Maps.

Implemented caching with gzip compression and LUT remapping to avoid per-frame reprojection of ~2.17M grid cells.

Offloaded GRIB2 decoding to Web Workers and OffscreenCanvas for smooth UI during large dataset processing.

Client-side NEXRAD Level II decoder handling Bzip2 decompression and binary parsing for all seven radar moments.

WebGL-based polar-to-Mercator projection using custom vertex shaders for smooth pan/zoom rendering.

Secured $5,000 in funding at the Iowa Innovation Challenge for a real-time, automated steel-tip dart scoring system.

Fine-tuned YOLOv4-tiny on 3,000+ annotated dart images with Raspberry Pi integration and mobile app sync.