The CS2 program funds research that elevates correctness—defined as the absence of faulty behaviors like numerical errors, data races, and memory faults—to a first-class requirement in scientific computing tools and workflows, on par with performance. It exists because the increasing complexity and scale of scientific simulations, data analysis, and emerging workflows (e.g., AI/ML, multi-physics) demand rigorous formal reasoning and verification to ensure trustworthy results, especially for DOE mission-critical and NSF broad-science applications.

- Intellectual Merit: How well does the proposed research advance the state of the art in correctness for scientific computing? Includes novelty, potential impact, and soundness of the formal methods or verification approach. - Broader Impacts: Benefits to society, education, workforce development, and scientific progress. For Eniola, an Africa/Nigeria angle (e.g., capacity building, open-source tools for LMIC researchers) could be a strong broader impact. - Collaboration Plan: Must demonstrate close, continuous collaboration between a scientific computing expert and a formal reasoning/verification expert. This is a mandatory requirement. - Eligibility: Only US academic institutions (universities, colleges) or eligible US non-profits can submit. The PI must be at a US institution. International collaborators are allowed but cannot be PI. - Budget Justification: Must include funds for collaborative activities (travel, workshops, personnel). - Data Management Plan: Must address reproducibility and correctness of data and code. - Postdoctoral Mentoring Plan: Required if postdocs are budgeted.

Past winners are typically US-based academic research groups (e.g., from MIT, Stanford, UC Berkeley, UIUC, University of Texas) with joint PIs from computer science (formal methods, PL, verification) and computational science (physics, biology, climate, engineering). Examples include projects on verified numerical libraries (e.g., using Coq or Isabelle for floating-point correctness), runtime verification for HPC simulations, and property-based testing for scientific workflows. No specific named examples were found on the page, but the archetype is a multi-PI team with a track record in both formal verification and a specific scientific domain.

A US-based tenured or research-track faculty member (or equivalent) with a strong publication record in both formal methods (e.g., theorem proving, static analysis, model checking) and a scientific computing domain (e.g., climate modeling, molecular dynamics, astrophysics). The ideal applicant has existing collaborations bridging these fields, a clear vision for how correctness can be proven or tested in a performant system, and a plan for training students in both areas.