Topic of the software project in WS25/26: LLM-Based News Agent

Project start: 15.10.2025

More information

The use of AI agent systems opens up a wide range of possibilities for the rapid and flexible development of intelligent applications. The use of large language models (LLMs) is particularly relevant in news processing, as they not only summarize content but can also be adapted to different contexts and styles. By combining AI agent frameworks with modern language models, students have the opportunity to apply key technologies in a practical way while simultaneously questioning their methodology.

In this project, students will build on an existing AI agent framework. The goal is to implement a news crawler that automatically captures current news from public sources. Using LLMs, the content is condensed into compact summaries and displayed in different journalistic styles. One focus is on adapting the language models to the task at hand, allowing students to test their configuration, fine-tuning, and, if necessary, fine-tuning. This will deepen the project both technically and methodologically. In addition, students research freely available datasets and suitable benchmarks to systematically evaluate the quality of the results. Optionally, sentiment analysis can be integrated to capture the emotional tone of news content.

The project is being conducted as a practice-oriented software project in cooperation with an external partner / customer. Students work in self-organized teams and implement an existing agent framework in a real-world application. They acquire in-depth knowledge in the practical use, adaptation, and fine-tuning of LLMs, learn methods for the systematic evaluation of text systems, and practice interdisciplinary skills such as project organization, teamwork, and scientific research. The open nature of the project encourages independent work and allows for subsequent continuation as part of a bachelor or master thesis.