Offene Thesis

Regardless of your career choice, bachelor's and master's theses are important projects that allow you to demonstrate your skills and research competencies. Our department offers numerous opportunities and we are always happy to welcome students who are interested in working with us!

Here are just a few examples of topics you could tackle in our department:

 

  • "Low Power AI": The energy consumption for training and inference of a Large Language Model (LLM) and models for Automatic Speech Recognition (ASR) is very high. The subject of the research is the implementation and comparison of various methods and techniques to reduce energy consumption in the application field of speech and text understanding. Possible approaches are to be discussed in the area of ​​knowledge distillation, factorization, quantization or neuromorphic computing.
  • "Chatbots": Many chatbot applications in the commercial environment require a "task-oriented dialogue", e.g. when reserving hotels, transferring money or reporting claims to the insurer. These dialogues are not very natural language based in the state of the art. The subject of the research is the implementation and comparison of various methods and techniques that enable natural language interaction and also take real-time information into account, e.g. recommendations for events in the region. Possible approaches are to be discussed in combination with a Large Language Model (LLM).
  • "Automatic Legal Text Processing": The automatic processing of legal texts is a challenge. The texts are semi-structured, linked in many ways and sometimes very extensive. The subject of the research is the implementation and comparison of various methods and techniques that enable the robust and comprehensible processing of legal texts. Possible approaches are to be discussed in the areas of statistical parsing, knowledge graphing and large language model (LLM).

  • "genAI for Code": With generative AI, programming has changed significantly and will change even further in the coming years. Writing the source code is giving way to a description in natural language or in graphical form. Checking source code is becoming more important, which is made very difficult by numerous security additions. The subject of the research is the generation of source code and information that facilitates checking.

     

If you are interested in working in our department, suggesting your own topic or simply want to find out more, please contact Professor Georges at Munir.Georges@thi.de 

The resulting software and results will be part of an open source solution, which will also be the basis for future scientific articles.

Please note that no work with a confidentiality notice will be supervised.