Stefano Markidis

Stefano Markidis's research addresses the computational demands of activities like daily weather forecasts, the development of future fusion nuclear reactors, and the training of large language models such as ChatGPT. These applications require extensive high-performance computing infrastructure, consisting of supercomputers with millions of computing units like CPUs and GPUs, along with specialized units such as reconfigurable hardware and tensor processing units, interconnected through a network. To meet the challenges of high-performance computations, diverse computing paradigms have been explored, including classical supercomputers, early quantum computers that utilize quantum mechanics principles, and neuromorphic systems designed to replicate the brain's functionality within a CPU. Markidis's work is centered on creating programming approaches and software specifically designed to facilitate applications like weather forecasting, fusion reactor simulations, and the training of large AI systems on extreme-scale computing systems. This includes not just traditional supercomputers but also innovative devices like quantum computers and neuromorphic hardware. His efforts aim to enable effective and efficient programming of large-scale systems and emerging computing platforms, making simulations a crucial tool for understanding and tackling significant challenges such as clean energy system design and climate change simulations.