LaTeX allows users to create complex layouts and large document structures efficiently, provided they possess the necessary expertise. However, editors at a Swiss publishing house are increasingly encountering LaTeX-related issues, particularly during the manuscript processing stages. Multiple points of interaction within the company's workflow involve LaTeX, and the growing number of annual submissions utilizing this tool has highlighted process inefficiencies. This thesis aimed to identify the main obstacles contributing to these inefficiencies and how they could be tackled.

This thesis focuses on developing a chatbot powered by a LLM to assist LaTeX users in troubleshooting issues independently. The project began with determining requirements and selecting the most suitable LLM for integration into the chatbot. Subsequently, the chatbot's performance was tested, evaluated, and refined using methods to enhance output quality.

The study found that, with the current state of technology and available resources, the LLM chatbot performed best when combined with prompt engineering. After deployment and testing, the chatbot demonstrated strong performance in addressing both simple and complex LaTeX-related queries. The final evaluation concluded that the chatbot was successful and could be integrated into the publishing house's workflow. This integration is expected to streamline processes and improve overall efficiency.