Capabilities

Algebraic and Eigen Solvers

FASTMath Contacts: Sherry Li and Rob Falgout

Data Analysis

Data analysis transforms large-scale simulation and experimental data into scientific insight. At DOE scale, analysis capabilities must operate efficiently on distributed and heterogeneous computing systems, often while data is still being generated. Expertise in data analysis spans in situ and streaming analytics, feature detection and tracking, topological analysis, statistical characterization, and production visualization. These methods enable researchers to identify important phenomena, quantify uncertainty, explore large parameter spaces, and accelerate scientific discovery while reducing the costs associated with storing and moving massive datasets. The RAPIDS portfolio provides capabilities in scalable scientific visualization, feature extraction, topological analysis, workflow-integrated analytics, and in situ processing. These technologies allow scientists to analyze data during execution, detect and track critical features, characterize complex behaviors, and interactively explore datasets that would otherwise be too large to manage using traditional post-processing approaches. The RAPIDS technology supports applications ranging from fusion energy and climate science to particle physics and materials research.

LEADS Contacts: Henry Kvinge, Stephen Young

FASTMath Contacts: Rick Archibald and Julie Bessac

RAPIDS Contacts: Rob Latham and Ana Gainaru

Data Management

Data management includes the method design, software, and infrastructure required to efficiently store, move, organize, access, and curate scientific data throughout its lifecycle. As simulations, instruments, and AI workflows produce increasingly large and distributed datasets, effective data management becomes essential for scientific discovery. Key capabilities include scalable I/O, data streaming, workflow orchestration, campaign data management, metadata services, data movement across storage hierarchies, and performance monitoring of data access patterns.The RAPIDS portfolio provides technologies for high-performance I/O, workflow-enabled data orchestration, campaign-scale data management, and understanding application data access behavior. These capabilities help scientific teams efficiently move data between memory, storage, analysis services, and AI workflows while maintaining scalability across leadership computing facilities. RAPIDS technology enables performance characterization and optimization of scientific data workflows, helping researchers reduce bottlenecks, improve resource utilization, and support increasingly integrated simulation, experimental, and AI-driven campaigns.

RAPIDS Contacts: Rob Latham and Ana Gainaru

Decision Support Methods

FASTMath Contacts: Bert Debusschere and Jeff Larson

LEADS Contacts: Guannan Zhang, Jerome Darbon

RAPIDS Contacts: Ana Gainaru

AI for Science

AI for Science combines artificial intelligence, advanced mathematics, scientific computing, and domain knowledge to accelerate scientific discoveries. It is especially important for DOE because the Department brings together leadership-class computing, major experimental facilities, advanced instruments, and mission-scale datasets. Connecting these assets through AI can create an integrated environment in which models, simulations, experiments, and instruments are coordinated in closed-loop workflows. This supports the Genesis Mission objective of increasing research productivity and shortening the path from scientific question to validated result and practical impact.

Our combined portfolio spans the scientific AI lifecycle. It includes physics-aware surrogate and reduced-order models that accelerate expensive simulations; generative, geometric, and graph-based methods for modeling complex systems; uncertainty quantification, inverse design, optimization, digital twins, and learning-enabled control. We also develop agentic workflows that can plan multistep investigations, invoke simulations and analysis tools, interpret results, and adapt subsequent actions. Another distinctive capability is training, evaluating, and deploying scientific foundation and multimodal models at leadership scale---including language, spatiotemporal, time-series, and multimodal models, with demonstrated campaigns on DOE leadership computing platforms.

FASTMath Contacts: Todd Munson and Bert Debusschere

LEADS Contacts: Panos Stinis and Guannan Zhang

RAPIDS Contacts: Shinjae Yoo and Sandeep Madireddy

Discretization

FASTMath Contacts: Mark Shephard and David Gardner

Application Performance

For the past 25 years, SciDAC's performance engineering and optimization efforts have translated the exponential increases in system performance spanning the tera-, peta-, and exascale eras into commensurate increases in application and workflow performance. This capability is delivered through a tight loop that integrates performance modeling and analytical tools, advances in communication libraries and runtimes, and partnership-focused performance engineering efforts that see our cohort of HPC experts immerse themselves in the complexity of each domain's computational science and seamlessly meld their decades of trustworthy performance engineering with the accelerated productivity of AI coding agents. Building decades of research and development in performance tools, performance optimization, performance and energy modeling and analysis, autotuning, etc., SciDAC experts develop and deploy techniques, technologies, and tools to enhance high performance, energy efficiency, portability, and productivity. To showcase and jump start partnership and community adoption of novel computer science, data, AI/ML, and applied math techniques, we can help develop proxy applications. Finally, we explore and develop core capabilities in the use and analysis of emerging AI and neuromorphic accelerators and concepts to address SciDAC partnership needs.

FASTMath Contacts: Cody Balos

LEADS Contacts: Miroslav Stoyanov and Guannan Zhang

RAPIDS Contacts: Sam Williams and Xingfu Wu

Software Quality and Sustainment

Software quality and sustainment encompass the engineering practices, metrics, and community investments that keep scientific software correct, reliable, and maintainable across its lifetime. As DOE science grows more dependent on complex, long-lived software running on rapidly evolving leadership-class architectures, deliberate attention to quality and sustainment is essential to protect the substantial investment these codes represent and to ensure that scientific results remain trustworthy and reproducible. Expertise in this area spans the adoption of software engineering best practices---version control, continuous integration, automated testing, documentation, and secure release processes---together with objective software metrics that make quality measurable, and the community development and training activities that grow and retain a skilled contributor base. These capabilities allow scientific teams to catch defects early, quantify and steadily improve the health of their codebases, and lower the barriers for new developers to contribute. The RAPIDS portfolio provides hands-on stewardship of widely used software products, helping teams establish CI testing, improve code coverage and static analysis, and adopt community standards such as OpenSSF scorecards. RAPIDS also develops shared, metric-based frameworks---assessing impact, sustainability, and quality---that let projects benchmark their practices and demonstrate progress over time. Through targeted mentoring, documentation, and training, RAPIDS helps DOE teams deliver dependable capabilities today while building the practices needed to sustain them well into the future.

FASTMath Contacts: Ann Almgren

LEADS Contacts: Miroslav Stoyanov and Jan Drgona

RAPIDS Contacts: Berk Geveci

Institues

There are currently three SciDAC institutes with over 24 participating institutions. The mission of the SciDAC institutes is to provide intellectual resources in applied mathematics and computer science, expertise in algorithms and methods, and scientific software tools to advance scientific discovery through modeling and simulation in areas of strategic importance to the US Department of Energy (DOE) and the DOE Office of Science (SC).

FASTMath — Frameworks, Algorithms, and Scalable Technologies for Mathematics

The FASTMath Institute develops and deploys scalable mathematical algorithms and software tools for reliable simulation of complex physical phenomena and collaborates with domain scientists to ensure the usefulness and applicability of FASTMath technologies.

Institute Director: Carol Woodward, Lawrence Livermore National Laboratory

DOE Program Manager: Xujing Davis

RAPIDS—SciDAC Institute for Computer Science and Data

The RAPIDS Institute solves computer science and data technical challenges for SciDAC and SC science teams, works directly with SC scientists and DOE facilities to adopt and support RAPIDS technologies, and coordinates with other DOE computer science and applied mathematics activities to maximize impact on SC science.

Institute Director: Rob Ross, Argonne National Laboratory

DOE Program Manager: Marco Fornari

LEADS—SciDAC Institute for LEarning-Accelerated Domain Science

LEADS introduces a paradigm shift by integrating SciML directly into domain-specific challenges. Working closely with domain scientists, the LEADS team will structure their SciML approach to domain science problems in various complexity levels and assign the proper SciML capability. LEADS is uniquely positioned to address phenomena where traditional numerical methods may be insufficient and expand DOE’s capability in using machine learning for domain science.

Institute Director: Panos Stinis, Pacific Northwest National Laboratory

DOE Program Manager: Xujing Davis

Last updated: 04 August 2026