The German Climate Computing Centre (DKRZ) is the central simulation and data processing facility for the German climate and Earth system modelling community and one of the leading institutions in this field worldwide. DKRZ not only operates supercomputers of the highest performance class and one of the largest data and archive systems, but also participates in many national and international projects aimed at improving the software and infrastructure for climate modelling.

Research Software Engineer (all genders) HPC & Performance Portability

Your tasks

A key focus for DKRZ is the development of the climate and weather model ICON, and its further preparation for current and future HPC systems. Within the European PHOENIX project, we aim to develop a maintainable, performance-portable version of ICON that runs efficiently on heterogeneous CPU and GPU architectures. In this context, you will contribute to the development, integration, and optimisation of ICON on the largest European supercomputers.

Your work will focus on several aspects of performance‑portable scientific software, depending on your background and interests. This includes integrating ICON components based on different programming approaches, such as C++/Kokkos, DaCe, and gt4py, improving their interoperability, and enabling efficient execution on different HPC architectures, particularly NVIDIA and AMD GPUs, but also possible future compute architectures.

Further areas of work include improving ICON’s efficiency and scalability at very high spatial resolutions. You may also explore the use of mixed precision and AI‑oriented hardware features for climate models, or improve the efficient concurrent execution and coupling of ICON with other components. The specific focus of your work will be determined together with the project team, allowing complementary contributions to the different development tasks.

Qualifications / Experiences

Academic degree in computer science, applied mathematics, computational physics/engineering, or a related field with a clear focus on scientific computing
Experience in high-performance computing, performance portability, and optimising scientific applications for modern hardware
Experience with scientific software development, ideally involving large, collaboratively developed applications
Programming experience with Fortran or C/C++, ideally in the context of scientific or parallel applications
Experience with parallel programming, for example MPI or GPU programming
Familiarity with performance‑portable programming approaches such as DaCe, gt4py, or similar technologies is an advantage
Programming experience under Unix/Linux, including scripting languages such as Python or Bash; experience with HPC environments, performance analysis, or code optimisation is an advantage
Motivation to support open, collaborative, and community‑oriented software development
Very good communication skills and knowledge of English

Employment conditions

We offer work in interdisciplinary teams and excellent connections to national and international research networks. In addition to specific development goals and requirements from our project partners, your work will be characterised by the freedom to explore and implement solutions at the interface between high-performance computing, scientific software engineering, and Earth system modelling. Participation in international conferences, project meetings, and hackathons provides opportunities for further training, exchange with the international HPC community, and presentation of your own results wherever meaningful.

We are based in Hamburg, but want to create a suitable working environment by offering flexible working hours and location (up to 100% work from home within Germany). Remuneration will be based on the candidate's qualifications according to TVöD-Bund. You can ideally start already in February 2027, but the exact starting date can be agreed individually. The position is financed by third‑party funded projects and is limited until the end of 2029.

Questions?

Dr. Jan Frederik Engels ( ) and Dr. Hendryk Bockelmann ( ) will be happy to answer them in advance.

#J-18808-Ljbffr

Deutsches Klima rechenzentrum GmbH