The German Climate Computing Center (DKRZ) is the central simulation and data processing facility for the German climate and Earth System modelling community and is one of the leading facilities in this area. DKRZ not only operates supercomputers in the highest performance class and one of the largest data and archive systems worldwide, we also participate in many national and international projects aiming to improve the software and infrastructure for climate modelling.

COMPUTER SCIENTIST / RESEARCH SOFTWARE ENGINEER FOR A LLM BASED CLIMATE DATA ANALYSIS FRAMEWORK (ALL GENDERS)

The Data Analysis department of the DKRZ is looking for a Computer Scientist / Research Software Engineer (all genders) to work on a Chat-Based Geoscientific Research Transformer (chatGRT). Together with our partner institution the Helmholtz Centre Potsdam GFZ funded by the Volkswagen Foundation “Pioneering Research”, we are building an AI-powered, chat-based tool that frees scientists from tedious data handling and coding tasks, allowing them to focus on groundbreaking discoveries. By seamlessly connecting top research tools with diverse climate data, chatGRT translates research questions into actionable scripts—delivering answers, plots, and insights at the click of a button.
DKRZ intends to hire an expert for software development around LLM and their applications around climate data analysis theme within the new Data Analysis (DA) department bridging the AI/ML to the Analysis Platform side (Freva) of DA. In this role, you will play a key part in revolutionizing how climate science is conducted by developing and fine-tuning chatGRT—an AI-powered, chat-based interface designed to assist Earth system scientists. You will work on both experimental and production aspects, from refining GPT-based prototypes to deploying an open-source large language model on a high-performance computing (HPC) infrastructure. Collaborating closely with domain scientists and AI experts, you will ensure that chatGRT is not only scientifically accurate and efficient but also secure, reproducible, and user-friendly.

QUALIFICATIONS:

· You have a master level degree in computer science, natural science or a comparable education with a focus on information technology. A PhD is a plus.
· Experience in AI/ML, its software components, and large language model architectures.
· Experience in high-performance computing (HPC) systems, software engineering, data workflows and scientific analysis.
· Strong programming skills (Python), especially in environments relevant to data handling and processing (CLI). Experience with scientific gateway technologies is a plus.
· Excellent communication skills in English; German is a plus.
· You are a team player with excellent problem-solving abilities and a keen interest in contributing to climate science research.

• Experimental Development:
• Fine-tune GPT-based prototypes to support geoscientific tasks such as annotated plotting, statistical analysis, and climate data processing. Develop and integrate experimental geoscientific research tools using existing models and datasets (e.g., ERA5), and facilitate user data uploads.
• Deployment & Integration:
• Implement and fine-tune open-source LLMs (e.g., DeepSeek, Llama, Gemma) for deployment on DKRZ’s HPC infrastructure. Ensure seamless integration with DKRZ services, including cloud storage and HPC job scheduling via SLURM. Integrate chatGRT seamlessly into the Freva analysis platform at DKRZ as FrevaGPT for the most convenient user experience.
• Benchmarking & Testing:
• Collaborate with domain scientists to develop benchmark tests covering common climate tasks (e.g., calculating the ENSO index) and HPC job script formulations. Monitor performance, speed, and scientific accuracy to iteratively refine the system.
• Security & Responsible AI:
• Develop a secure, responsible AI framework that restricts the model’s responses to scientific queries and prevents malicious advice. Establish pipelines for processing user feedback and maintaining the system through regular updates and monitoring.
• Documentation & Collaboration:
• Contribute to comprehensive documentation and ensure reproducibility across various programming and virtual environments. Engage with the broader scientific community by sharing prototypes on platforms such as GitHub and setup of workshops.