Three UMBC Information Systems students, Pavithra Gottipati, Andrew Meiller, and Sakshi Jaju teamed up in IS 636 – Structured Systems Analysis and Design, taught by IS Professor Lei Zhang, to design and prototype an AI-powered chatbot for the myUMBC portal. Their project focused on improving how undergraduate and graduate students find answers to questions and get support about campus services.
Their chatbot supports topics including billing, financial services, library resources, career services, and IT support; integrates with myUMBC Help/Request Tracker and Student Business Services; and includes analytics, monitoring, and administrative tools. They used a Retrieval-Augmented Generation (RAG) approach with
- A sentence transformer model retrieves the closest frequently asked question (FAQ) using vector embeddings
- A Microsoft Phi-3 Mini Instruct LLM system to generate fallback responses when no close match is found in the FAQ
- Secure NLP/LLM hosting, integrated knowledge base, and API connectivity within the UMBC ecosystem
Read more about their project here.