Skip to content
Jangara Bliss
All projects
Product & EntrepreneurshipSelected workAug 2024 – Oct 2025

KDUR Community Radio Data Platform

A deployed data platform for Fort Lewis College's radio station — 60+ daily users — with an applied-AI research layer for search and royalty integrity.

Role — Research assistant — built and deployed the platform; prototyped the AI layer

Microsoft Power AppsPythonVector embeddingsNeo4jCypher
Research poster
Problem
The station's catalog, schedules, and analytics lived in scattered records — and royalty attribution breaks when the same artist appears under five name variants.
System type
Deployed CRUD platform + applied-AI prototype
Why it matters
Sixty people using your software every day is a different education than any class project: uptime, usability, and data integrity stop being abstract.
Team context
Research assistantship under Dr. Matthew Welz, Fort Lewis College.
The deployed KDUR radio station management application
The deployed station-management app.

01

Overview

Under Dr. Matthew Welz: a complete database application built in Microsoft Power Apps that manages KDUR's music library and broadcast schedules for over 60 daily users — real software, in production, with real stakeholders. On top of the deployed platform sits a research layer: a vector-embedding model in Python that resolves artist-name variants (matching "Jay Z" to the official "Jay-Z") so royalty attribution stays accurate, a Neo4j graph-database prototype, and AI agents that translate natural language into Cypher queries.

System architecture

DJs and staff work in the deployed Power Apps platform; the research layer sits alongside it — embeddings for entity resolution, a graph prototype for relationship queries, and natural-language-to-Cypher agents for conversational access.

  1. DJs & staff — 60+ daily users
  2. Power Apps platform — library + schedules
  3. Embedding layer — artist-name resolution
  4. Neo4j prototype — NL → Cypher agents
KDUR platform architecture: users, Power Apps CRUD application, and the applied-AI research layer
Platform + research layer.
Neo4j graph database prototype visualization of the KDUR catalog
Neo4j prototype of catalog relationships.

02

Contributions

  • Developed and deployed the full Power Apps database application for the music library and broadcast schedules — in daily production use by 60+ users.
  • Built a vector-embedding model in Python for artist-name resolution to protect royalty attribution accuracy.
  • Prototyped a Neo4j graph database and AI agents translating natural language into Cypher queries — toward conversational access to the catalog for students and community.

03

Evidence & evaluation

Evidence

Production deployment

attached

In daily use by 60+ station users.

Platform screenshots

attached

Deployed app and graph prototype in the gallery.

Entity-resolution evaluation

pending

Quantify match accuracy on a labeled variant set.

Metrics

Daily users

60+

Resolution accuracy

Not yet measured

Evaluate the embedding matcher before quoting figures.

04

Limitations

  • The AI layer is a research prototype — the embeddings and NL-to-Cypher agents aren't yet part of the production deployment.

05

Lessons & tradeoffs

  • Real users reorder your priorities immediately — reliability and data integrity beat features every week.
  • [Add the specific user-feedback moment that changed the design.]to fill

06

Artifacts