Retrieval Augmented Generation (RAG) for university administrative documents

This project develops a prototype Retrieval-Augmented Generation (RAG) pipeline for automated answering of recurring queries from university administrative documents. Source texts will be preprocessed, split into semantically meaningful chunks, and stored in a vector database via pretrained embedding models. A retriever will then be evaluated on a manually curated test set and paired with a large language model (e.g. GPT-4 or LLaMA2) to generate transparent, explainable answers. The goal is to demonstrate RAG’s value in university administration and reveal its potential to streamline communication.