Vývoj a nasazení studijního asistenta GenAI do produkčního prostředí Vysoké školy ekonomické v Praze

Generative artificial intelligence offers educational institutions new ways to provide students with access to information. However, deploying it safely and reliably in a real university environment is a complex engineering challenge. This thesis documents the development and deployment of a study assistant system based on a Retrieval-Augmented Generation (RAG) architecture at the Prague University of Economics and Business, as part of the Study Buddy project. Development is structured into three phases: a Proof of Concept validating the basic feasibility of the RAG pipeline over study materials, a Minimum Viable Product migrating the solution to the Microsoft Azure cloud platform, and a production solution including automated transformation and upload pipelines together with a governance framework. The thesis addresses three research questions: what architecture and technology combination in the cloud environment is optimal for the given problem, what are the key challenges and best practices when transitioning a GenAI prototype to production at a public university, and how the quality and structure of input metadata affects RAG system performance. The result is a working study assistant that answers only based on verified study materials, is integrated into the VŠE infrastructure, and keeps internal data within a controlled environment. The thesis also delivers a data management framework, a specification of operational requirements for continuous development, and an institutional governance framework. It also serves as a case study transferable to other institutions planning similar generative artificial intelligence deployments.