We are the Helmholtz AI consultants for the research field energy located at Karlsruhe Institute of Technology.
This page provides an overview of our work, highlighting our

• 🧑‍🔬 research interests,
• 💻 key software packages, and
• 🗒️ most significant publications.

We’re thrilled to have you here and hope you find what you're looking for 🙌.

Helmholtz AI is an application-driven AI platform within the Helmholtz Association, Germany's largest scientific organization. Designed to bridge the gap between traditional scientific research and AI, we aim to integrate AI into the Helmholtz Association's core research areas. Our mission is to democratize AI by connecting AI researchers with domain experts and ensuring access to resources and expertise, making the latest advancements in AI accessible to everyone. The annual Helmholtz AI project call is an attractive funding initiative aimed at supporting collaborative, interdisciplinary AI research projects within Helmholtz. Our dedicated HPC system HAICORE (Helmholtz AI Computing Resources) offers powerful computational resources specifically designed for AI applications, allowing Helmholtz researchers to perform large-scale data processing, training of ML models, and other compute-heavy AI tasks.

With Helmholtz AI consulting, we empower all Helmholtz researchers to leverage cutting-edge AI methods for solving their scientific problems. Our consultants provide support and expertise in applied AI, tools, and software engineering, ensuring that researchers in fields like biology, physics, or climate science can use AI technologies without needing to be AI specialists themselves. Working with our consultants comes at no cost, as collaborations are entirely scientific.

➡️ Check out our voucher system to set up your collaboration today!

Helmholtz energy researchers are working to find innovative solutions to meet the energy demands of current and future generations. As the Helmholtz AI consultant team @ KIT led by Markus Götz, we support the Energy research field by offering expertise in cutting-edge AI methods. AI applications in energy research are as diverse as the field itself, ranging from energy system load forecasting and discovering new materials for energy storage technologies like batteries, to automating industrial system control.

To address these challenges, we leverage advanced AI approaches with a special focus on:

• 🖼️ Image analysis
• 📈 Time series
• Graph-based problems
• Uncertainty quantification
• Model search
• Large-scale parallel processing

These capabilities allow us to tackle a variety of AI tasks, including regression, classification, segmentation, and interpolation. We are also committed to open and reproducible research, ensuring our code and data are accessible through open-source platforms.

To support our research and enable cutting-edge AI applications in the energy domain and beyond, we develop and maintain several key software packages and general purpose tools, each designed to address specific challenges in the field:

• 🍃 perun, a Python package that calculates the energy consumption of Python scripts by sampling usage statistics from your Intel, Nvidia, or AMD hardware components. It can handle MPI applications, gather data from hundreds of nodes, and accumulate it efficiently. perun can be used as a command-line tool or as a function decorator in Python scripts.

  J.P. Gutiérrez Hermosillo Muriedas et al. (2023). perun: Benchmarking Energy Consumption of High-Performance Computing Applications. In: J. Cano et al. (eds) Euro-Par 2023: Parallel Processing. Euro-Par 2023. Lecture Notes in Computer Science, vol 14100. Springer, Cham. https://doi.org/10.1007/978-3-031-39698-4_2

• 🧬 Propulate, an asynchronous population-based optimization algorithm and Python package for global optimization and hyperparameter search on supercomputers.

  O. Taubert et al. (2023). Massively Parallel Genetic Optimization Through Asynchronous Propagation of Populations. In: A. Bhatele et al. (eds) High Performance Computing. ISC High Performance 2023. Lecture Notes in Computer Science, vol 13948. Springer, Cham. https://doi.org/10.1007/978-3-031-32041-5_6

• 🌍 HyDe, the first GPU-accelerated hyperspectral denoising toolbox in Python.

  D. Coquelin et al. (2022). Hyde: The First Open-Source, Python-Based, GPU-Accelerated Hyperspectral Denoising Package. In: 2022 12th Workshop on Hyperspectral Imaging and Signal Processing: Evolution in Remote Sensing (WHISPERS), Rome, Italy, pp. 1-5. https://doi.org/10.1109/WHISPERS56178.2022.9955088

• 📈 ReCycle, a Python package for fast and efficient long time series forecasting with residual cyclic transformers.

  A. Weyrauch et al. (2024). ReCycle: Fast and Efficient Long Time Series Forecasting with Residual Cyclic Transformers. In: 2024 IEEE Conference on Artificial Intelligence (CAI), Singapore, Singapore, pp. 1187-1194. https://doi.org/10.1109/CAI59869.2024.00212