This thesis explores the use of multimodal large language models (LLMs) to improve the grading of UML diagrams and provide assistance to students in software engineering courses. The research focuses on two main areas: developing and enhancing an automated grading system for UML diagrams and creating a preliminary feedback system to support students while they create and refine their diagrams. By implementing these features within an existing learning management system, this thesis aims to address challenges related to grading consistency, tutor workload, and timely student support in large-scale university courses. The project will evaluate different approaches for automated grading, comparing image-based and text-based assessments of UML diagrams.