معرفی
Dr. Carolin Penke is a researcher at the Jülich Supercomputing Centre (JSC) of Forschungszentrum Jülich GmbH. She leads projects in training large language models (LLMs) on high-performance computing (HPC) systems as part of the OpenGPT-X initiative and the Accelerating Devices Lab.
- Current focus: HPC-AI integration, memory-efficient LLM training, and accelerator evaluation
- Prior work: numerical linear algebra and mathematical algorithm development for quantum chemistry
Her research spans computational mathematics, artificial intelligence, and high-performance computing. Key contributions include:
- Developing CARAML for AI workload evaluation
- Creating JUPITER benchmark suite for exascale systems
- Optimizing low-rank representations in deep learning
- Advancing pseudosymmetric matrix algorithms for physics simulations
- Designing GPU-accelerated numerical methods for eigenvalue problems
Her work bridges mathematical rigor with practical AI/HPC applications, emphasizing scalability, energy efficiency, and European sovereignty in AI.
۰مقاله منتشرشده
