Aegis Scholar: A Lightweight Network-Centric Tool for Defense Expertise Discovery

Author(s):

Mark A. Tschopp, Ethan Shafer, Steve McCoy, Adriana Canedo, Markuz Robinson, Jonathan Hale, and James R. Uplinger

Abstract:

This report documents the design, implementation, and evaluation of Aegis Scholar, a prototype expert-discovery platform developed under the AI Technicians program at the Army Artificial Intelligence Integration Center. Addressing the challenge of identifying subject-matter experts across distributed Department of War organizations, Aegis Scholar acts as a “connection engine” by integrating semantic vector search (Milvus) with relational graph mapping (Neo4j). By ingesting metadata from the Defense Technical Information Center, the system represents authors via abstract-derived embeddings and ranks them against natural language queries using a composite relevance heuristic. Offline evaluations demonstrated the efficacy of limited-context transformer models for semantic matching, while a task-based user study showed that the targeted, network-centric interface reduced search times by up to 50% for defense-specific queries compared with general-purpose tools. The project successfully demonstrates the operational value of minimalist, visual discovery interfaces for defense applications, while simultaneously serving as a robust process development vehicle for mid-career Soldiers executing end-to-end data science projects under ARL mentorship.