Scientific Researcher (f/m/d) AI-Based Replacement of Chemistry-Climate Simulations with a Mathem...

Helmholtz Association of German Research Centres; Deutschland; Vor Ort; EUR 40.000 - 60.000; Teilzeit

Zusammenfassung

A prominent research institution in Germany is looking for a candidate for a part-time scientific role focusing on AI application in Earth System Sciences. You will develop AI models to replace traditional simulations, analyze results, and contribute to mathematical understanding and optimization. This position offers a chance to pursue a doctoral degree while gaining valuable research experience.

Qualifikationen

• Experience with AI model architectures such as RNNs, CNNs, and Transformers.
• Strong background in optimization theory and statistical learning.
• Capable of conducting comparative analyses and simulations.

Aufgaben

• Develop concepts to replace traditional ESM simulations with AI models.
• Train AI models using datasets from long-term simulations.
• Conduct theoretical analysis on hyperparameter optimization.

Kenntnisse

AI model development

Dimensionality reduction

Hyperparameter tuning

Statistical learning theory

Mathematical optimization

Ausbildung

Master's degree in a relevant field

Tools

RNNs

CNNs

Transformers

PINNs

Jobbeschreibung

Area of research:

Other

Part‑Time Suitability:

The position is suitable for part‑time employment.

Starting date:

29.10.2025

Job description:

The Scientific Computing Centre (SCC) is a central scientific institution at KIT that performs tasks related to research, teaching, and innovation and provides comprehensive services within KIT and to external parties.

This research project focuses on the application of AI models in Earth System Science (efficiently replacement of a chemistry-climate simulation from an Earth System Model (ESM) by an AI-based approach) and the development of mathematically rigorous methodes for interpretable AI modeling and systematic hyperparameter optimization. In detail:

• Development of a concept to replace the chemistry climate simulation of the ESM with a suitable AI model. This includes testing and selecting appropriate AI architectures (e.g., RNNs, CNNs, PINNs, Transformers), identifying relevant input features, applying dimensionality reduction techniques, and performing hyperparameter tuning (learning rate, number of layers, regularization strength).
• Training the AI model using a dataset from long-term ESM simulations (from 1979 to 2024), following by running simulations with the trained AI model.
• Comparative analysis of AI‑based simulations versus traditional ESM simulations, assessing accuracy and performance.
• Moving beyond a black‑box approach, the project aims to achieve a mathematical understanding of the AI model. This understanding should be used to develop systematic methods for optimizing the hyperparameter tuning, grounded in the mathematical areas of optimization theory, statistical learning theory, and explainable AI.
• The theoretical analysis of this hyperparameter optimization - such as studying convergence properties, regularization effects, and sensitivity analysis - constitutes the core mathematical challenge of the project.

This position offers the opportunity to earn a doctoral degree while working in the CSMM research group led by Prof. Dr. Martin Frank and being part of the KIT's KCDS graduate school.

This research center is part of the Helmholtz Association of German Research Centers. With more than 42,000 employees and an annual budget of over € 5 billion, the Helmholtz Association is Germany's largest scientific organisation.