NASA
Solicitation: NNH25ZDA001N-RADIANT
Released Date: December 18, 2025
Required Letter of Intent Due Date: January 26, 2026
Proposal Due Date: April 27, 2026
RADIANT seeks innovative adaptations and applications of existing AI techniques, concepts, methodologies, etc. to demonstrate their feasibility and potential to increase science return, create AI ready data and data products as well as to inform Heliophysics science research disciplines of promising techniques and capabilities. A high priority is on those which interface to the Surya foundation model and demonstrate its value in accelerating scientific investigations.
RADIANT is intended to enable heliophysics science research to adopt AI tools to perform common functions. Notional areas of interest for proposals include, but are not limited to:
• Validation of the Surya foundation model through the use of a wide range of downstream applications.
• Demonstrations of the use of Surya in conducting scientific investigations.
• Tools and methods for analysis of observational Earth and heliophysics datasets to characterize solar impact on Earth’s climate change.
• Tools and methods for analysis of observational Earth and heliophysics datasets to identify correlations between solar and terrestrial phenomena (i.e., earthquakes, tsunami, etc.).
• Tools and methods to enable analysis of large volumes of heliophysics observational data to support solar and inner heliospheric processes.
• Tools and methods to enable analysis of large volumes of heliophysics observational data to validate independently an heliophysics model.
• Mature to the point of re-usability heliophysics-oriented capabilities previously developed under LSW/T&M, MDRAIT, HITS and/or other program elements.
• Demonstrate an approach for science validation of an AI model already developed. Proposed techniques must not use any of the AI techniques in the original model development.
• Tools that accelerate existing algorithms using ML surrogates like speeding up existing sub-components of physical models, ML-based PDE solvers, and whole physical model surrogates where there is already abundant input/output datasets available.
• Explainability & sensitivity for hypothesis testing and insight: Variational inputs to test hypotheses on processes that lead to target science outcome, varying training sets to produce an ensemble of models, thus ensemble of results, to estimate uncertainty with non-Gaussian or multi-modal outcomes, ML
architectures that explicitly include interpretable aspects.
• Create a comprehensive set of labeled features from an existing observational data set;
• Create AI data science products to advance automated mission operations;
• Develop general AI/ML data-related methods and approaches applicable to multiple NASA missions;
• Improving the calibration of data or using new techniques to calibrate data like self-calibration, etc.