2026 ASEE Annual Conference & Exposition

Multi-agent orchestration provides a systematic approach to distributed intelligence generation using specialized processing components. This implementation presents an agent architecture that comprises database retrieval, graph-based collaborative filtering, and multi-source content fusion agents, all of which are coordinated through intent classification routing. The system integrates structured course catalogs (MySQL), user interaction graphs (Neo4j), and vectorized knowledge repositories (FAISS) to demonstrate the coordination of heterogeneous data sources. We implemented intent classification using large language models to dynamically route queries to appropriate specialized agents based on semantic analysis. The architecture was validated using the IMPEL educational domain with real user interactions and course recommendations. The database agent handles direct catalog queries, the collaborative agent performs graph traversal for peer-based recommendations using 1024-dimensional user embeddings, and the content agent integrates resume parsing with real-time market intelligence through web search APIs. Results demonstrate successful query routing and response generation across diverse user intents, with each agent maintaining specialized functionality while contributing to unified system outputs. The modular design enables systematic extension to other domains that require distributed intelligence coordination, providing a practical framework for implementing multi-agent systems in production environments.