Logo Universität Bayreuth

PhD position (m/f/d) for the research area of Decentralized Machine Learning

Chair of Data Systems

At the University of Bayreuth, Faculty of Mathematics, Physics and Computer Science, at the Chair of Data Systems (Prof. Dr. Ruben Mayer), starting at 01. June 2024, we offer a

PhD position (m/f/d)for the research area of Decentralized Machine Learning

in full time.

About us:

The Chair of Data Systems at the University of Bayreuth, led by Prof. Ruben Mayer, is dedicated to performing high-quality research in the area of large-scale distributed systems for data management, especially in the area of machine learning (ML). We aim to solve real-world problems with data at scale.

Requirements:

  • Master’s degree in computer science (or related) with very good results
  • Interest in topics around the area of machine learning, distributed systems, and data management
  • Basic knowledge in distributed systems and machine learning
  • Experience with the development of software systems, very good skills in programming with standard programming languages such as C++, Python or Java
  • Experience with fundamental software engineering concepts such as version control, testing and debugging, CI/CD pipelines, etc.
  • Hand-on experience with machine learning frameworks (e.g., PyTorch) is a plus
  • Excellent command of English
  • Very good writing skills
  • High engagement, high motivation, pro-active communication skills, and high social skills

Your tasks:

Machine learning (ML) has enabled a breakthrough in the development of artificial intelligence in recent years. Centralized ML systems collect large amounts of data, e.g. from the Internet, and use it to train their models (see ChatGPT, LLAMA, etc.). However, only data that is publicly accessible can flow into the models. In many areas, data protection prevents data access, e.g. for medical and other personal data. To make such private data usable for model training, decentralized machine learning (also known as "federated learning") relies on leaving the data with the data sources (the "clients"). Instead of sharing the data, a local model is trained on the local data on each client and then aggregated on a round-by-round basis to obtain a global model. 

There are a lot of challenges to consider. Decentralized training scales less well, as the model parameters must be communicated by many clients via slow (Internet) connections. The distributed architecture enables new attack vectors. In addition, it is unclear how decentralized machine learning complies with AI regulations (especially GDPR and EU AI Act).

The aim of our research is to advance decentralized machine learning in this area and establish it as a powerful alternative to centralized machine learning. To this end, we are conducting research together with our partners (including IBM Research and the University of Cambridge).

We are looking for:

Did we spawn your interest to work in this exciting field and contribute to the progress of science?

To perform this exciting research, we are looking for a highly motivated student who has recently finished / is soon going to finish his/her master’s degree in computer science (or similar). The position is fixed-term according to the “Wissenschaftszeitvertragsgesetz” and serves the own scientific qualification (doctorate) with remuneration according to pay group 13 TV-L. In all other respects, the hiring requirements according to the Bavarian University Innovation Act (BayHIG) apply.

We offer:

  • Work in a highly innovative environment
  • Supervision at one of the leading universities of Germany
  • Funds for travel and student assistants (HiWis) are available
  • We will consider all incoming applications until the position is filled
  • as well as other benefits of public service, such as JobRad, attractive additional pension scheme with the Federal and State Pension Institution, a wide range of health promotion programmes, e.g. health days, workshops, active break and the opportunity to participate in a very large number of different sports programmes as part of the general university sports programme
  • In addition to holiday leave and the option of flexible working hours, time off on 24 December and 31 December of each year

    The University of Bayreuth appreciates the diversity of its employees as an enrichment and is expressly committed to the goal of equal opportunities for all genders. Women are strongly encouraged to apply. Applicants with children are very welcome. The University of Bayreuth is a member of the Best Practice Club “Familie in der Hochschule e.V.” and has successfully participated in the HRK audit “Internationalization of the University”. Persons with severe disabilities will be given preferential consideration if they are equally qualified.


Your application

Please apply via the online application portal of the University of Bayreuth online with keyword “Data-Systems”.
The documents will be deleted after the position has been filled in accordance with data protection requirements.

Required documents: Curriculum vitae, a short letter of motivation, if applicable a list of publications (also blog posts and software projects) and complete transcripts of your Bachelor and Master studies.

If you have any questions, please feel free to contact Prof. Mayer: ruben.mayer@uni-bayreuth.de

We look forward to receiving your application.