University AI Chatbot

University support staff were fielding repetitive queries about enrollment deadlines, academic policies, course requirements, and administrative procedures. High query volume during enrollment periods overwhelmed the support team, leading to long response times and student frustration. Many queries had answers already documented in policy documents and FAQs — but students couldn't find them. The institution needed a way to scale support availability without proportionally scaling headcount.

We built a retrieval-augmented generation (RAG) chatbot using Rasa as the conversation management framework, Ollama for running an open-source LLM locally (avoiding external API costs and data privacy concerns), and a vector database to store and retrieve relevant institutional documents. When a student asks a question, the system retrieves the most relevant policy documents, course information, or procedure guides from the vector store and feeds them to the language model as context for generating an accurate response. The chatbot is integrated into the university's web portal.

The chatbot now handles the majority of routine student queries autonomously, significantly reducing the load on support staff during peak periods. Students receive answers immediately at any hour. The local LLM deployment keeps all student data within the institution's infrastructure, satisfying data privacy requirements.