Projects

Problem

Manual RFP searching required analysts to spend months reviewing thousands of government contracts to find relevant fiber and telecom opportunities.

Solution

Built a production desktop application that applies Azure OpenAI (Azure AI Foundry) to automate web grounding, intelligent classification, and matching of RFPs to company capabilities. Features batch search, URL validation, and CSV export.

Impact

• Processing time reduced from months to hours
• Automated classification of thousands of contracts
• Deployed as a production tool at Olsson

Problem

Evaluating where to build new fiber networks meant analysts drawing polygons by eye, estimating passings per mile in spreadsheets, and recalculating every time a boundary changed — slow, inconsistent, and hard to scale.

Solution

Built a custom ArcGIS Pro decision-support add-in that reads address points and route networks, divides the study area into a spatial grid for density analysis, flood-fills viable build zones against configurable PPM thresholds, and generates economic scorecards. An interactive spot-check panel lets analysts refine boundaries grid-cell by grid-cell with live scorecard updates.

Impact

• Expansion planning reduced from days or weeks to minutes
• Consistent, repeatable candidate build areas across analysts
• Human-in-the-loop refinement keeps strategy in analyst hands

Problem

Analysts spent hours in browser portals — pick a state, wait for downloads, unzip, import, fix projections, repeat — before any real GIS work could start.

Solution

Built a custom ArcGIS Pro dock-pane add-in (.NET 8 / C#) that pulls address points, parcels, building footprints, OSM roads, TIGER boundaries, FEMA flood zones, USGS elevation, GBIF species records, Wikipedia markers, and BSL broadband data directly into the project geodatabase. Large SQL parcel pulls use auto-tiled extents, parallel streaming from SQL Server into GDB InsertCursor, and direct WKT-to-geometry parsing.

Impact

• Eliminates manual portal hunting and import loops
• Statewide parcel downloads at millions of rows without choking SQL
• Packaged as .esriAddinX — double-click install, no Visual Studio required

Problem

Checking poles and street-level equipment meant leaving ArcGIS Pro, copying coordinates into a browser, and hunting for the right Street View panorama — slow and easy to lose map context.

Solution

Built an ArcGIS Pro add-in (C# / .NET 8 / WPF) that calls the Google Street View API from a map tool and right-click action, rendering the matching panorama in a docked WebView2 panel. Optional second click sets viewing direction so analysts can verify assets without leaving the map.

Impact

• One-click street-level verification inside ArcGIS Pro
• Powered by the Google Street View API
• Distributed as a self-contained .esriAddinX installer

Problem

Finding where development is happening for fiber expansion meant manually reviewing huge parcel-owner tables — too slow to flag investment and development firms at scale.

Solution

Built a Python desktop app (Tkinter, PyInstaller) that applies Azure OpenAI to classify property owners as development/investment firms in SQL Server. Supports parallel batch workers, rate limiting and retries, keyword pre-filtering, confidence thresholds, and a live progress log.

Impact

• Applied Azure AI across large parcel datasets
• Faster fiber expansion analysis where development is concentrated
• Minimal manual review with tunable confidence and pre-filters

Problem

Laying out FTTH networks in ArcGIS meant days of manual digitizing — placing junction shafts, connecting every address, and designing a backbone — with frequent errors and no consistent drop-type or trunk classification for cost estimates.

Solution

Built an ArcGIS Pro add-in (.NET 8 / WPF) that automates the full workflow in three steps: place shafts at road intersections, connect homes to the nearest shaft via a spatial grid (tagging Street Drop vs Driveway Drop by geodesic length), then run Kruskal’s minimum spanning tree for an optimal backbone classified as Main Trunk or Terminal Branch. Respects active map selections for partial study areas.

Impact

• Network layout reduced from days to minutes
• Drop-type tagging drives material cost estimates automatically
• MST backbone minimizes total cable length for construction phasing

Problem

Wireless network planning meant juggling spreadsheets, specialized RF software, and GIS separately — hours of manual PCI assignment, interference checks, and coverage analysis with no shared map view for the team.

Solution

Built a custom ArcGIS Pro dock-pane add-in that reads standard tower CSV exports and runs eight capabilities in one place: antenna sector pie wedges on the map, automatic PCI assignment, azimuth and tilt optimization, signal-strength heatmaps (Okumura-Hata, COST-231, and other propagation models), color-coded co-channel and adjacent-channel interference lines between tower pairs, and site placement optimization using elevation, line-of-sight, and Fresnel zone diffraction — all written back as shareable map layers.

Impact

• RF planning workflows reduced from hours to a few clicks
• No context-switching between GIS, spreadsheets, and external RF tools
• Map-native outputs the whole engineering team can review together

Problem

Identifying utility poles, streetlights, and telecom infrastructure in aerial imagery required expensive manual inspection across large geographic areas.

Solution

Built a desktop application that fetches high-resolution aerial tiles by coordinate, runs a custom-trained YOLO model to detect utility assets, and exports georeferenced results to ArcGIS. Supports single-point, grid, and area scanning modes with tile navigation controls.

Impact

• Automated detection of utility poles and streetlights from aerial tiles
• Georeferenced export to GIS-ready formats
• Custom YOLO model trained on telecom infrastructure classes

Problem

Field surveys for utility poles and telecom equipment require costly in-person visits, and traditional GIS methods cannot identify specific physical assets from ground level.

Solution

Built a desktop application that traverses streets using street-level imagery, running a custom YOLO model to detect and classify utility infrastructure. Supports line traversal, panoramic capture, and area buffering with step-by-step navigation controls.

Impact

• Remote field survey capability without site visits
• Automated inventory of streetside utility assets
• Configurable traversal distance and panoramic heading coverage

Problem

Generating bore profiles for fiber network directional drilling required days of manual drafting, creating bottlenecks in project timelines and introducing human error.

Solution

Developed a fully automated desktop application that reads spatial waypoints from an interactive map, processes elevation models, calculates bore depth and slope, and generates production-ready 2D and 3D elevation profiles with configurable running lines and break points.

Impact

• Processing time cut from days to minutes
• Eliminated manual drafting errors
• Deployed across the telecom engineering team at Olsson

Problem

Complex fiber network design workflows required repetitive manual GIS operations, slowing down engineers and creating inconsistencies across projects.

Solution

Developed a suite of custom Python geoprocessing toolboxes for ArcGIS Pro that automate routing analysis, cost estimation, and network design workflows.

Impact

• 90% reduction in manual processing steps
• Accelerated fiber network design timelines
• Improved cost estimation accuracy

Problem

Limited understanding of spatiotemporal patterns of NOx emissions from U.S. cement plants and their environmental justice implications.

Solution

Used TROPOMI satellite data to perform spatiotemporal hotspot analysis of NOx emissions, correlating emission patterns with population demographics and environmental justice indicators.

Impact

• Full MS thesis — GPA 4.00
• Remote sensing analysis of all U.S. cement plants
• Environmental justice exposure analysis for affected populations

Problem

Historical patterns of racial exclusion in Omaha real estate were undocumented spatially, making it difficult to understand the geographic scope of discriminatory practices.

Solution

Digitized historical land parcels from archival documents and aerial photography, georeferenced historical maps, and built a spatial database revealing redlining and racial covenant patterns.

Impact

• Revealed spatial patterns of racial exclusion
• Contributed to urban spatial justice research
• Accurate historical parcel database for Omaha