Research Field: Software Engineering

This master’s thesis aims to enhance educational outcomes on Artemis by reducing tutor workload and improving the student learning experience. We’ll build on Athena’s preliminary work that deployed Large Language Models (LLMs) like GPT-4 to provide automatic feedback on text-based exercises. The focus is to employ more advanced LLM techniques such as Retrieval Augmented Generation (RAG) and fine-tuning Llama 3 with historical feedback data. The project will proceed in two phases: refining the feedback system for more context-sensitive, automated responses, and then evaluating and comparing the system’s accuracy, efficiency, and educational impact.

Apollon is an online modeling editor written in React. It offers an easy to use and lightweight experience to learn UML and other notations in an interactive environment. It runs as standalone version on https://apollon.ase.in.tum.de offering a collaborative modeling experience without accounts. It is also integrated into Artemis to enable modeling exercises. Students create models, tutors assess those models and provide feedback which students can then see in order to improve their modeling skills.

Programming exercises in Artemis can be realized with Bitbucket/Bamboo, Gitlab/Jenkins, GitlabCI or LocalVCS/CI. This allows flexibility for Artemis admins. LocalVCS/CI provides an easy setup for developers and for universities to try out Artemis. However, not all functionality for programming exercises is available on all combinations of version control and continuous integration subsystems In this thesis, the missing features should be implemented. In addition, the concept of build agents and dockerized builds should be realized within LocalCI.

Artemis supports any programming language for programming exercises. Currently, it includes templates for many popular languages, however templates for a couple of languages is still missing. In this thesis, new templates for programming exercises will be developed and tested with the Bitbucket/Bamboo and Gitlab/Jenkins and LocalVCS/CI. In addition, static code analysis should be integrated into more programming exercise templates. Artemis is open source and available on https://github.com/ls1intum/Artemis

The number of computer science students is rising significantly each semester. To meet the demand for prompt and effective feedback, particularly in programming tasks, it is essential to enhance Athena, a platform for (semi-)automated text and programming assessments, to facilitate a higher volume of (semi-)automated feedback. This thesis aims to develop and integrate two new modules into Athena, evaluating them based on resource consumption, including computational capacity and memory usage, and the relevance of their feedback to specific programming tasks.

This thesis contributes to the enhancement of the Team Allocator for Software Engineering courses, an open-source tool for forming student teams. The focus is on optimizing data import and export functionalities, refining the matching algorithm, improving the user interface and experience, and implementing a serverside component to enable synchronous multi-user collaboration. The methodology employed is agile development with 2-3 week product increments, allowing for regular testing and release cycles. The main contribution of this work lies in providing significant upgrades to the TEASE project, enhancing its usability and efficiency in project-based educational settings.

This thesis proposal addresses communication challenges in Artemis. The aim is to enhance Artemis by introducing a refined communication module. The identified problems include duplication in communication sections, ab- sence of document-sharing, and limitations in threaded conversations. Moti- vated by the need for a robust communication platform, this thesis proposal suggests improvements guided by Nielsen’s Heuristics. Key objectives involve a unified messaging system, enhanced message thread functionality, collabo- rative document sharing, and improved UI design.

Getting an overview of different REST calls and the respective endpoints of large software systems is a tedious and error-prone task, as there are many endpoints, and the generation of URIs is often profoundly nested. Faulty endpoints may lead to decreased usability of a system. Manually analyzing the client- and server-sided authorization steps is equally tedious, as it happens in many places. Faulty authorization may have severe consequences. This thesis describes the development of a GitHub action analyzing client-sided REST calls and the respective server-sided REST endpoints, the used URLs, and the client- and server-sided authorization.

The rapid evolution of information and communication technologies led to widespread adoption of E-learning systems, particularly during the pandemic when traditional in-person education was not possible. This shift underlined the importance of diverse learning rhythms and paces among students, a challenge particularly in larger classes. To address this, Artemis, an open-source interactive learning platform, introduced learning paths tailored to individual learner goals and capacities. However, despite the functionality of learning paths, Artemis currently lacks an intuitive user interface, hindering the adoption and usability of the feature.

This master’s thesis aims to refine an LLM-based feedback system for program- ming assignments in the Artemis Learning Management System. Despite previous advancements, challenges remain in providing accurate, adaptive, and scalable feedback. The proposed enhancements include the integration of model vector databases, specialized models like Code Llama 2, and an agentic-based approach to address issues of language compatibility and feedback precision. The objective is to reduce manual workload for educators and provide immediate, tailored feed- back to students, thereby enhancing the educational experience and efficiency in learning programming skills.

The emergence of Kubernetes has transformed the practices surrounding the deployment and management of containerized applications by introducing advanced automation, scalability, and operational ease. This thesis focuses on enhancing the Kubernetes deployment at the Technical University of Munich. We identify key issues within the current deployment architecture, including the lack of persistent storage in the Kubernetes cluster, making it unsuitable for stateful applications like Artemis, the cumbersome quality assurance workflow for Artemis, and lacking access control hindering educational use of the Kubernetes cluster.

Artemis is an online learning platform that offers students and instructors a wide range of interactive learning experiences. However, the platform contains design inconsistencies as well as confusing and complex layout elements in the exam mode, course overview and tutorial groups from the student’s point of view. As a result, the complexity and distractiveness of the layouts hamper the user ex- perience, making it difficult for students to navigate and utilize the platform ef- fectively.

Artemis is a learning management platform that offers students different exercise types as well as lecture materials. To help students navigate the exercises and lectures Artemis offers learning paths and makes suggestions about the order of the resources. This thesis aims to optimize the learning path recommendations within Artemis by integrating repeated testing and improving the accuracy and transparency of progress metrics. The expected outcome is a more efficient and engaging learning experience for students, with personalized suggestions that adapt to their learning pace and needs.

The interactive learning platform Artemis offers students a unique opportunity to learn and train software engineering skills online. Especially for advanced courses and topics, students need to set up a local integrated development environment (IDE) to work on their exercises which can be a daunting task. The main objectives of the thesis include the integration of the customized online IDE Theia within Artemis and the setup of a performant execution environment on Kubernetes.

Personalized learning experiences are pivotal in enhancing student engagement and academic success. This thesis investigates the development of dynamic learner profiles within Artemis, an open-source Learning Management System (LMS). By leveraging data analytics and machine learning techniques, we aim to generate comprehensive learner profiles that capture individual learning behaviors, preferences, and progress. These profiles will enable tailored educational experiences, providing insights for both educators and learners. The study covers the methodology for data collection and analysis, the design and implementation of the profiling system, and the impact of personalized learning on student outcomes.

In the rapidly evolving landscape of educational technology, scalable and maintainable software architectures are crucial for the success of learning management systems (LMS). This thesis explores the transformation of Artemis, an open-source LMS, from a monolithic architecture into a modular system using Java Modules. By modularizing Artemis, we aim to enhance the system’s scalability, maintainability, and development efficiency. The thesis delves into the challenges of decomposing a large monolithic codebase, the strategies employed for effective module separation, and the benefits realized post-modularization.