AI for Science

(Interim) Lead: Shantenu Jha, Head - Computational Sciences

Objective: The AI For Science Group harnesses the transformative potential of artificial intelligence (AI) to advance scientific research. The group aims to accelerate discoveries, enhance data analysis and develop innovative solutions to complex problems by integrating AI into scientific workflows. This interdisciplinary initiative opens new avenues across domains beyond fusion and plasma physics, including materials science, electromanufacturing, and aerosol science. The group enables advances in computer science, data science, and translational research across these areas. Our expertise includes: (1) Developing sophisticated algorithms, enhancing machine learning techniques, and ensuring seamless integration of AI tools into existing scientific frameworks; (2) Applying these advances for predictive modeling, data-driven hypothesis generation, and automation of experimental Computational Sciences Department design; (3) Delivering and tailoring these capabilities and AI solutions to address specific scientific challenges end-to-end and at scale.

Applied and Computational Mathematics

Lead: Ammar Hakim, Deputy Head - Computational Sciences Department

Objective: The Applied and Computational Mathematics (ACM) group develops mathematical descriptions, models, and algorithms to describe complex systems, often involving a vast range of time and/or spatial scales. The groups core focus is to design new algorithms for multiscale, multi-physics problems that are of interest across the Department of Energy (DOE) complex and elsewhere. Our key areas of expertise include computational plasma physics, computational fluid dynamics, design of higher-order methods for solution of partial differential equations (PDEs), particle (Lagrangian) methods, new asymptotic methods, and analysis of numerical algorithms. A key new thrust is using machine learning for PDEs' forward and inverse solutions. Our group also develops production-quality software, incorporating these methods and making them available to the larger DOE and scientific community. Our recent focus has been on developing new Explainable AI driven surrogate and solver development, focusing on bringing in advanced tools from mathematics and theoretical computer science to bear on the problem of accelerating time-to-solution.

Computational Partnerships and Technologies

Lead: Stephane Ethier, Principal Computational Scientist, Computational Sciences Department

Objective: The Computational Partnerships and Technologies group consists of multidisciplinary scientists who advance computational methods and architectures to domain applications. This involves creating new ways to simulate and analyze data, as well as improving the systems — such as supercomputers and specialized software — that make advanced computing possible. Expertise in implementing new numerical methods, high-

performance computing, performance optimization using software engineering best practices, and new architectures are applied to experimental and computational projects. The team collaborates with computational scientists at universities, other DOE Labs, and private companies to use our methods to produce problems and applications.