You will work in the HClimRep project, which is funded by Germany’s Helmholtz Association and aims to develop a pioneering AI foundation model for ocean and climate research. This advanced model will integrate concepts from foundational models used in speech and image processing. It will be designed for a range of tasks including seasonal to decadal climate predictions, “what-if” experiments, and comprehensive ocean and atmosphere modeling. The project, which involves a consortium including the Forschungszentrum Jülich, Alfred Wegener Institute, Karlsruhe Institute of Technology, and Helmholtzzentrum Hereon, utilizes a deep learning model with several billion parameters, drawing on data from various sources, including world-leading physics-based climate models.
Tasks
You will be part of a team responsible for preparing the training data, including data from existing model runs with the ocean and climate models used at AWI. It will also entail running data assimilation experiments by optimally combining the models and observations. You will also work together with the team at Forschungszentrum Jülich in developing and training the foundation models with several billion parameters on the high-performance computing (HPC) facilities at Jülich Supercomputing Center, including JUPITER—Europe’s first exascale HPC system. Finally, you will also be involved in the evaluation of the foundation models, with emphasis on assessing the model’s ability to represent important physical processes and phenomena and consequently extrapolate.
Requirements
- A master’s in climate science, atmospheric physics, oceanography, physics, informatics, or a related field is required. Having a Ph.D. would be a distinct advantage.
- Demonstrated skill/proficiency in working with big data is a further requirement.
- In-depth knowledge of at least one programming language (e.g., Python, R or Fortran) is also expected.
- Furthermore, excellent English skills, both written and spoken, are expected.
- Given that the success of this project hinges on effective collaboration between partners, strong communication skills and the ability to work in teams are both a necessity.
Having some of the following skills would be a distinct advantage:
- Experience with AI (e.g., transformer-encoder stacks)
- Experience with GPU-based HPC systems
- Good understanding of climate physics
- Hands-on experience with climate models
- Track record of publishing research in peer-reviewed journals
- Ability to make good presentations (e.g., at major conferences).
Application deadline: July 31, 2024
