UniversitasAI
An AI Operating System for higher education utilizing 25 active SphereAgents and Chain-of-Provenance metadata to eliminate AI hallucinations.
Project Overview & Objectives
UniversitasAI represents the next evolution of academic planning and institutional intelligence. Traditional university environments rely on massive, disjointed legacy databases like Banner, PeopleSoft, and custom SQL warehouses to manage class scheduling, course requirements, research metadata, and student records. Administrators and advisors spend hundreds of hours manually auditing schedule conflicts and compliance data. UniversitasAI unifies these systems into a real-time autonomous agent network. Operating 25 role-based SphereAgents, the system automatically crawls databases, resolves scheduling overlaps, and answers complex administrative queries. To guarantee institutional compliance with FERPA and local government mandates, the platform incorporates a strict Chain-of-Provenance Retrieval Augmented Generation (RAG) framework, ensuring every AI-generated response is backed by a verifiable database source citation.
Core Tech Stack
- SphereAgents
- NextJS
- Neo4j Graph
- Provenance RAG
- Node API
The Challenge
Higher education institutions operate with fragmented legacy systems (e.g., Banner, PeopleSoft) that isolate academic, financial, scheduling, and student success data. This fragmentation makes tracking performance and administrative planning slow, while standard commercial AI engines risk student data leaks and hallmarked inaccuracies.
Our Solution
Tilda designed and built UniversitasAI, an autonomous institutional operating system. Leveraging 25 specialized, role-based SphereAgents and 106 autonomous data scanners, the system maps and crawls across legacy databases. Using a zero-hallucination Chain-of-Provenance RAG architecture, it outputs secure, verified answers in full compliance with FERPA regulations.
Interactive Device Preview
Data Flow & Pipeline Architecture
A technical overview of how data is ingested, processed, and routed through the system.
Federated Scan
Autonomous database scanners constantly crawl Banner/PeopleSoft databases looking for updates.
SphereAgent Broker
Orchestrates course conflicts, financial calculations, and student queries across specialized agents.
Provenance Verification
Verifies security access clearances and generates responses with explicit citation mapping.
Development & Delivery Timeline
Key phases of the development sprint from initial database design to production rollout.
ERP Integrations & Database Schemas
Establishing federated read-replicas and writing GraphQL/REST connectors for legacy campus data.
SphereAgent Multi-Agent Orchestration
Setting up the agent broker network, task queues, and conflict-resolution matrix.
Chain-of-Provenance RAG Implementation
Fine-tuning vector embedding models on pgvector and integrating FERPA-compliant encryption layers.
Sandboxing & Security Framework
Every digital product we engineer complies with modern security frameworks and operating system boundaries.
SphereAgent Coordination
Orchestrates multi-agent pipelines to execute cross-portal administrative and scheduling actions.
Chain-of-Provenance RAG
Maintains verifiable source-attribution records for all generated queries, eliminating hallucinations.
Federated ERP Integration
Secure connectors for legacy Banner, Workday, and PeopleSoft institutional databases.