Master Thesis (f/m/x) - Implementation of an OKF-based LLM chatbot for the Aerospace Industry

The DLR Institute of Software Technology sees software as a catalyst for research and innovation. The institute's staff, currently numbering around 200, make a significant contribution to advancements in the fields of aviation, space, energy, transportation, and security through the development of state-of-the-art software solutions and innovative research. Our areas of competence include reliable and safety-critical software systems, artificial intelligence, high-performance computing and quantum computing, human-system interaction and visualisation, software and systems engineering as well as digital platforms and digital twins.

What To Expect

In the Flight Software Department, we conduct research and develop reliable and resilient (real-time) software for aircraft and spacecraft. With our expertise in this field, we currently support approximately 20 space missions with software engineering and software quality assurance.

This thesis aims to explore the state of the art technologies in Large Language Models, with a focus on investigating the way alternative knowledge representations can improve the capabilities of retrieval-augmented generation algorithms. In current chatbot implementations, capturing the deep semantic structures of extensive technical documentation is a challenge. The project aims to investigate the potential of the newly proposed Open Knowledge Format (OKF) as a framework which preserves the hierarchical structure of domain-specific information. The research investigates opportunities where an AI-aided approach to knowledge management may improve upon classical methods.

Your tasks

• overview of current state-of-the-art methods to employ LLM technologies for specialized domains
• overview of robust evaluation methodologies for chatbot performance
• conducting a literature review on OKF and a comparison with RAG systems
• design and implement a framework based on current the state of the art (OKF with or without RAG) that focuses both on speed and accuracy
• introduce a mechanism for evaluating the system on response time, precision and other possible metrics
• perform analysis of the outcomes comparing traditional RAG systems with the OKF approach, on very structured and hierarchical input documents
• create an AI-powered compliance assistant that streamlines ECSS documentation navigation, serving as an important and reliable reference tool to accelerate decision-making for aerospace professionals

Your profile

• enrolled in a relevant master program (computer science, computer engineering, AI engineering, robotics, informatics, or a closely related field) and looking for a thesis topic
• programming languages (e.g., C/C++, Python, Bash)
• familiarity with LLM technologies
• familiar with Git, unit testing, CI/CD basics, and documentation
• able to communicate in English, as it is the supervision language

We look forward to getting to know you!

If you have any questions about this position (Vacancy-ID 5201) please contact:

Alexandros Marantos

Tel.: +49 8153 28 1493