Student assistant (f/m/x) - Pruning of Large Language Models

The DLR Institute of Data Science in Jena focuses on finding solutions to the new challenges of the digitalization era. Research concentrates on the areas of data management, data analysis and data acquisition. Three departments have been set up in line with the thematic focus of the institute.

What To Expect

In the Data Analysis and Intelligence department, methods are developed and applied that enable the analysis of complex and large data sets. Here, methods of machine learning, causal inference and domain-specific process knowledge are used. To increase the technology transfer potential, human factors such as acceptance are taken into account during application development.

In the "Machine Learning" working group, we research and develop innovative data-driven methods for data analysis and, in cooperation with other DLR institutes, find solutions for practical applications using machine learning methods.

As part of our working group, we offer you the opportunity to write your thesis or work as a student in the field of machine learning, specifically on the topic of pruning for large language models. As part of an international team, your task will be to help design, implement and evaluate new methods.

Your tasks

• study and understand with the state-of-the-art architectures of open-source large language models, like Llama, Gemma, Phi, etc.
• literature survey on various model compression methods, such as pruning, sparsity-based pruning, etc.
• realization and prototypical implementation of own or existing pruning methods in Python (or similar)
• scientific evaluation and interpretation of the results and comparison with existing methods
• test and documentation of the work
  
  

As part of your student work, you will be responsible for up to 20 hours of work per week. The exact topic of a thesis will be defined together with you according to your specific qualifications and expectations.

Your profile

• ongoing studies in computer science, mathematics, physics, data science or related subjects
• practical programming skills (Python, or similar; experience with frameworks such as Tensorflow/Keras or PyTorch )
• very good knowledge of English or very good knowledge of German
• knowledge of machine learning, deep learning, data analysis and evaluation
• basic knowledge of mathematics (analysis, linear algebra, stochastics, logic)
  
  

We look forward to getting to know you!

If you have any questions about this position (Vacancy-ID 1716) please contact:

Julia Fligge-Niebling

Tel.: +49 3641 30960 152