Scientific Machine Learning for Modeling and Simulations | DE FOA 0002319

Opportunity Information: Apply for DE FOA 0002319

The Department of Energy (DOE) Office of Science released this discretionary grant funding opportunity, titled "Scientific Machine Learning for Modeling and Simulations" (Funding Opportunity Number: DE-FOA-0002319), to push forward research at the intersection of machine learning, high-performance computing, and scientific simulation. The central idea is that scientific machine learning (often described as Scientific AI/ML) can be trained on scientific data to augment or automate tasks that traditionally required extensive human expertise, and that DOE is in a unique position to accelerate progress because it already supports large scientific datasets, powerful computing and networking platforms, and a national research workforce.

The opportunity is motivated by a DOE Basic Research Needs workshop report (2019) that laid out six "Priority Research Directions" (PRDs) for Scientific AI/ML. Three of these are foundational themes that cut across most scientific ML work: making methods domain-aware (so models respect scientific context and constraints), interpretable (so outputs can be understood, trusted, and interrogated), and robust (so methods behave reliably under noise, uncertainty, distribution shifts, and other real-world complications). The remaining PRDs are framed around major capability use cases: analyzing massive scientific datasets (PRD #4), enhancing modeling and simulation with ML (PRD #5), and enabling intelligent automation and decision support for complex systems (PRD #6). This FOA is primarily focused on PRD #5, meaning it prioritizes research that uses ML to improve, accelerate, or otherwise strengthen scientific modeling and simulation workflows.

A key theme in the description is that DOE science has long relied on complex, resource-intensive numerical simulations, sometimes characterized as "inner-loop" modeling (traditional forward simulations repeatedly executed inside larger analysis, optimization, or uncertainty-quantification loops). The FOA argues that combining established scientific computing approaches with ML-based adaptivity and acceleration could substantially increase the speed, throughput, and practical usefulness of these inner-loop simulations. At the same time, it emphasizes that ML methods themselves must be scalable and efficient to work with the massive datasets and high-performance computing (HPC) environments common in DOE science, which implies a strong interest in algorithms and software that can effectively use large-scale parallelism and modern supercomputing architectures.

The solicitation also highlights practical computing challenges that are becoming more important over time, especially the "extreme heterogeneity" of emerging scientific computing technologies, such as diverse processor types, complex memory hierarchies, high-speed interconnects, and increasingly integrated sensor and edge components. Because future performance gains depend on using these heterogeneous systems well, DOE signals interest in new techniques, tools, and software approaches that can extract scientific value from this evolving hardware landscape rather than being bottlenecked by it.

Another notable aspect is the explicit call for deeper engagement from the computational mathematics community. The FOA points out that scientific computing has decades of experience in numerical analysis and algorithm development, and that this expertise can improve the "inner loop" of ML training itself (for example, through better optimization methods, stability analysis, error control, and principled approaches to uncertainty). In other words, DOE is not only looking for applied ML demonstrations, but also for work that strengthens the mathematical and statistical underpinnings of predictive AI/ML models and adaptive algorithms used in scientific settings.

From an administrative standpoint, the opportunity was offered as a grant under the Science and Technology and other Research and Development activity category (CFDA: 81.049). Eligibility is listed as unrestricted (open to any type of entity, subject to any additional eligibility language in the full announcement). The FOA was created on April 8, 2020, with an original closing date of May 29, 2020. DOE anticipated making around 25 awards, with an award ceiling of $150,000 per award.

• The Department of Energy - Office of Science, Office of Science in the science and technology and other research and development sector is offering a public funding opportunity titled "Scientific Machine Learning for Modeling and Simulations" and is now available to receive applicants.
• Interested and eligible applicants and submit their applications by referencing the CFDA number(s): 81.049.
• This funding opportunity was created on Apr 08, 2020.
• Applicants must submit their applications by May 29, 2020. (Agency may still review applications by suitable applicants for the remaining/unused allocated funding in 2026.)
• Each selected applicant is eligible to receive up to $150,000.00 in funding.
• The number of recipients for this funding is limited to 25 candidate(s).
• Eligible applicants include: Unrestricted (i.e., open to any type of entity above), subject to any clarification in text field entitled Additional Information on Eligibility.

Apply for DE FOA 0002319

[Watch] Creating a grant proposal using the step-by-step wizard inside the applicant portal: