ATML Extreme-Scale AI and Simulation (exAIS)
The ATML Extreme-Scale AI and Simulation (exAIS) focuses on engineering algorithms, artificial intelligence (AI) workloads, and applications for scalability on the largest supercomputers and disruptive computing technologies. This requires design and development activities along the complete software hierarchy: from basic numerical methods optimized for the most recent hardware architectures over problem-specific algorithms to application-tailored workflows for domain scientists. In consequence, the lab brings together expertise form hardware architecture, numerical methods, data science, software engineering, and scientific applications. A particularity of the ATML is its collaboration with the respective domain scientists from leading academic institutions like CERN, RWTH Aachen, TU Munich, ETH Zurich, Barcelona Supercomputing Center, Riken, Stanford University, and industry partners like NVIDIA, Siemens, and Airbus.
With the goal of answering the most relevant scientific questions and enabling collaborating researchers worldwide to succeed, we leverage the JUPITER supercomputer for training large-scale AI models and running extreme-scale simulations. Furthermore, we lead both the JUPITER Research and Early Access Program (JUREAP) and the JUPITER AI Factory (JAIF), two of the most visible projects in European HPC and AI.
Research Topics
- High-scaling numerical methods
- Multimodal AI
- Complex workflows enabling extreme-scale applications
- AI-approaches for reduced order models
- Applications in various domains: Energy, health, climate, physics, CFD, …
