Jared Alexander Dumale

Early-warning workflows in intelligence shops still bottleneck at the moment when an analyst has to read, extract, and geocode unstructured text. The question I set out to answer: could a small LLM-driven pipeline absorb that load and hand back something an analyst could actually act on?

I built a proof-of-concept that ingested open-source reporting, used a language model to extract named entities and locations, generated structured event summaries, and pushed geocoded results into a GIS dashboard for trend visualization. The pipeline ran end-to-end without manual cleanup steps.

Twelve demo sessions with analysts produced concrete feedback on accuracy, presentation, and trust calibration — the kind of input that turns a research prototype into a deployable tool. The work demonstrates how applied GeoAI can compress the time between signal and decision.

USAFA cadets need to internalize which countries fall under which Combatant Command — but the rote method (flashcards, slides) doesn't stick, and the testing format rewards short-term memorization over geographic intuition.

Geo412 reframes the problem as a game. It's a browser-based platform with multiple trivia modes built around the world map, designed for self-paced study that rewards quick recognition and pattern association. Cadets play in short sessions; the geographic structure of the answers reinforces itself.

The project demonstrates that small, focused web tools can outperform institutional training material when designed around the actual cognitive task. It's live, used, and shows how curriculum design and lightweight development reinforce each other.

COVEY 535 was a USAF aircraft lost over Laos. Decades on, recovery teams still need defensible search prioritization — and the available data is fragmented across historical imagery, terrain models, and archived mission records.

I led an undergraduate research team that consolidated 22 historical imagery scenes and DEM-based terrain analyses into a single geospatial workflow. The output: four map products that refined priority search zones, fused from three distinct data types, and developed in coordination with three institutional partners.

This project illustrates how geospatial analysis can serve a humanitarian mission while developing student capability simultaneously. It's a case study in integrating mentorship, multi-source fusion, and decision-support cartography on a problem with a real outcome.

Career planning at USAFA depends on understanding the training pipeline for each Air Force Specialty Code: where you go, when, and for how long. The institutional view of this data is fragmented across PDFs, briefings, and informal advisor knowledge.

The AFSC Atlas — a capstone I supervised — consolidates that data into a single ArcGIS Dashboard. A cadet enters their AFSC and immediately sees training base locations on a map, sequenced milestones on a timeline, and key duration figures in plain view. The underlying dataset was curated and structured to be reusable beyond this proof-of-concept.

Beyond the product, the project demonstrates how a small geospatial dashboard can transform an advisor/cadet conversation from speculation to specifics — and how supervising student work can produce institutional artifacts the academy can actually use.