NORMAN: A Retrieval Artificial Intelligence Augmented Teaching Assistant for Higher Education

Conference Paper

M. Donovan, S. Shadangi, D. Ganskop, C. Beaton, R. Mahajan

Rochester Institute of Technology, United States

EDULEARN26 Proceedings — 18th International Conference on Education and New Learning Technologies
Palma, Spain · June 29–July 1, 2026

Norman is an AI teaching assistant built to address the limitations general-purpose language models run into in the classroom—hallucination, lack of source grounding, and no awareness of institutional access controls. It combines retrieval-augmented generation (RAG), role-based access control, and traceable source attribution over faculty course materials, running offline on the Qwen3.5 foundation model with course content stored as vector embeddings in ChromaDB to keep answers accurate and institutional data private.

RAG LLMs ChromaDB Human-Centered Design

DOI: 10.21125/edulearn.2026.0944  ·  ISBN: 978-84-09-88444-5  ·  ISSN: 2340-1117

View in IATED Digital Library →