Architecture for LLM-based e-learning system

Project description

The goal of the project was designing an architecture that enables use of LLMs in e-learning environments while accounting for privacy and security concerns.

Context

The research proposes an architecture that supports use of LLMs in e-learning systems. Use-case for validating the research is CY2 and an onboarding tool implemented as a Canvas module. It is meant to be deployed and integrated within the system. The proposed architecture provides an API end-point to which a front-end of choice can be connected.

Results

Most important product of the research is the proposed architecture which also addresses individualized learning approach and common security concerns. The architecture is explained, argumented and alternatives are presented along with possible uses and scaling options. Main proposed architecture was validated by applying it to CY2 use case and creating a proof of concept. Further theoretical validation was performed using ISO25010 model for software system quality. Security concerns and risks are addressed through OWASP Top 10 for LLMs and AI risk management framework.

About the project group

Previously finished bachelor of applied IT at Fontys, with a focus on cyber security and graduation in software.