Two AI surfaces — Natural-language Search and Community Insights — over a shared entity-typed RAG layer, kept current by three orchestrated autonomous agents with sub-agents.
The hardest thing to scale at Tech Square is the same thing that makes it valuable: dense colocation of people, ideas, and partnerships. Without a platform, that value lives in calendars and hallway conversations.
In an innovation community,who collaborates with whom, on what, funded by whomis the value. We retrieve over structured fields with weighted scoring, then ground LLM reasoning in the matched evidence — preserving the identity and relationships that vector-only RAG flattens.
People, University, and Organization agents each coordinate sub-agents fetching, verifying, and updating continuously. AI as operational substrate, not a feature.
Every answer shows the path to it. Each claim stays tied to the records it drew from — grounded in matched evidence, not asserted.
Search shows reasoning beside the score. Community Insights cites the entities behind every claim. The AI's work stays visible.
Plain-language member intents return ranked candidates with per-match reasoning.
Natural-language Q&A with citations to the underlying entities.
Coordinated sub-agents enrich profiles continuously from LinkedIn, publications, grants, and web sources.
Crawls faculty directories and fans out per-faculty enrichment to sub-agents
Tracks corporate-member footprint and innovation activity via coordinated sub-agents
The platform was assembled using the following technologies
