Show all projects

All projects
Optimizing Aircraft Performance for Enhanced Efficiency and Reduced Environmental Impact
Brandon M. Lowe Scale-Resolving Methods for Accurately Predicting the Aerodynamics of In-Flight Icing
David A. Craig Penner Enabling Sky-High Commutes: Simulating NASA’s Tilt-Wing Air Taxi Concept
David Garcia Perez Scale-Resolving Simulations for Jet Noise Prediction
Olivia Martin Supercomputing in Supersonics: Analyzing Noise Predictions from High-Speed Aircraft
Marian Nemec Parallel Geometry-Based Mesh Adaptation for Supersonic Jet Acoustic Analysis
Chase Ashby
Reliable Thermal Protection Systems for Planetary Missions
Savio Poovathingal Analyzing the Artemis I Launch Environment with Multiphase CFD
Chris DeGrendele SpaceX Starship Hot Staging - From Concept to Flight
Jason Liu Decoding Martian Dust: Brownout Simulations of the Ingenuity Mars Helicopter
Patricia Ventura Diaz Coupled Simulations of Propellant and Vehicle Dynamics Over Extreme Duration Time Scales
Marco Sansone
NASA-GPT: Searching the Entire NASA Technical Reports Server Using AI
Derek Dalle NASA’s OCSDO: Creating a Culture of Innovation and Collaboration
Andrew Mitchell Data-Driven Movies from the NASA Scientific Visualization Studio
AJ Christensen NASA's Newest Hybrid-Chip Systems
Reid Ormseth, Brian Roth Inspire – Next-Generation HPC at NASA Goddard
Robert Budden, Jonathan Mills HEC: Meeting Today’s Most Challenging Problems and Preparing for the Future
William Thigpen
Solar Corona Live! Predicting the 2024 Eclipse in Real Time
Ronald Caplan To Infinity and Beyond! NASA Plunges into a Black Hole
Jeremy Schnittman Forecasting Space Weather Using Machine Learning
Spyros Kasapis Unveiling Neutron Stars: Simulations and AI Insights into Cosmic Emissions
Constantinos Kalapotharakos, Greg Olmschenk The Sun in Action! 3D Realistic Modeling from Tiny to Large Scales
Irina Kitiashvili
Applying for NASA Supercomputing Internships
Laura Carriere How NASA Uses Filmmaking Tools for Scientific Data Visualization
AJ Christensen Introduction to NASA and Early Career Pathways
Matt Bruzek Computational Methods for Evaluating Transonic Aircraft Buffet
Daniel Maldonado Supercomputing in Supersonics: Quiet Supersonic Flight
Marian Nemec Becoming a NASA Intern and Exploring the World of CFD Simulations
Patricia Ventura Diaz How NASA Manages User Access Using Arbiter2
Dave Chau Managing and Patching Multiple Architectures at NASA
Don Story Designing Ansible for Large Scale Environments
Alex Pelkey

AI Foundation Models for Science

Our Planet
Figure illustrating true vs. predicted (non-dimensionalized) momentum fluxes in the upper troposphere (12-kilometer height).
From the Prithvi WxC foundation model’s gravity wave flux parameterization, the figure illustrates true vs. predicted (non-dimensionalized) momentum fluxes in the upper troposphere (12-kilometer height). Panels (a) and (c) show the fluxes u′ω′ and v′ω′ from the ERA5 reanalysis, and panels (b) and (d) show the respective predictions from the fine-tuned model. All fluxes are monthly averaged for May 2015. The vertical derivative of the fluxes represents the wind-forcing tendencies due to atmospheric gravity waves and can represent a portion of unresolved sub-grid tendencies in climate models. Aman Gupta, Stanford University; Sujit Roy, NASA/Marshall

The NASA Office of the Chief Science Data Officer, in collaboration with the five science divisions, is developing a suite of artificial intelligence (AI) foundation models tailored to the needs of NASA science. An AI foundation model is a large machine learning model pretrained on vast datasets using self-supervision, enabling it to perform a wide range of tasks with minimal additional training. This initiative is highly relevant for NASA’s Science Mission Directorate (SMD), as it maximizes the potential of NASA’s archive of science datasets, reduces the effort required to build AI applications, simplifies usage of scientific data, accelerates discovery, improves data-driven decision-making, and helps uncover new insights across datasets.

Building AI foundation models requires substantial amounts of computational resources with the latest GPUs. GPUs from NASA High-End Computing services and consulting expertise were crucial to pretrain the 2.3 billion-parameter Prithvi WxC foundation model.

Rahul Ramachandran,
NASA Marshall Space Flight Center

Quick Facts

  • Researchers used extensive GPU resources at the NASA Advanced Supercomputing (NAS) facility to pretrain the AI foundation model for Weather and Climate (Prithvi WxC). The pretraining process used 16 nodes (totaling 64 A100 GPUs) and ran for approximately 15 days.
  • Prithvi WxC uses NASA MERRA-2 reanalysis datasets generated at the NASA Center for Climate Simulation and has three distinct downstream applications: hurricane track forecasting, downscaling of weather prediction and climate projection, and a gravity waves parameterization.
  • NASA expects that the science community will build additional downstream applications using the Prithvi WxC foundation model.
  • The goals are to continue to improve the model and build additional foundation models for other disciplines such as heliophysics.

Researchers

  • Rahul Ramachandran, NASA Marshall Space Flight Center
  • Manil Maskey, NASA Marshall Space Flight Center

More Information

All projects

Optimizing Aircraft Performance for Enhanced Efficiency and Reduced Environmental Impact
Brandon M. Lowe Scale-Resolving Methods for Accurately Predicting the Aerodynamics of In-Flight Icing
David A. Craig Penner Enabling Sky-High Commutes: Simulating NASA’s Tilt-Wing Air Taxi Concept
David Garcia Perez Scale-Resolving Simulations for Jet Noise Prediction
Olivia Martin Supercomputing in Supersonics: Analyzing Noise Predictions from High-Speed Aircraft
Marian Nemec Parallel Geometry-Based Mesh Adaptation for Supersonic Jet Acoustic Analysis
Chase Ashby
Reliable Thermal Protection Systems for Planetary Missions
Savio Poovathingal Analyzing the Artemis I Launch Environment with Multiphase CFD
Chris DeGrendele SpaceX Starship Hot Staging - From Concept to Flight
Jason Liu Decoding Martian Dust: Brownout Simulations of the Ingenuity Mars Helicopter
Patricia Ventura Diaz Coupled Simulations of Propellant and Vehicle Dynamics Over Extreme Duration Time Scales
Marco Sansone
NASA-GPT: Searching the Entire NASA Technical Reports Server Using AI
Derek Dalle NASA’s OCSDO: Creating a Culture of Innovation and Collaboration
Andrew Mitchell Data-Driven Movies from the NASA Scientific Visualization Studio
AJ Christensen NASA's Newest Hybrid-Chip Systems
Reid Ormseth, Brian Roth Inspire – Next-Generation HPC at NASA Goddard
Robert Budden, Jonathan Mills HEC: Meeting Today’s Most Challenging Problems and Preparing for the Future
William Thigpen
Solar Corona Live! Predicting the 2024 Eclipse in Real Time
Ronald Caplan To Infinity and Beyond! NASA Plunges into a Black Hole
Jeremy Schnittman Forecasting Space Weather Using Machine Learning
Spyros Kasapis Unveiling Neutron Stars: Simulations and AI Insights into Cosmic Emissions
Constantinos Kalapotharakos, Greg Olmschenk The Sun in Action! 3D Realistic Modeling from Tiny to Large Scales
Irina Kitiashvili
Applying for NASA Supercomputing Internships
Laura Carriere How NASA Uses Filmmaking Tools for Scientific Data Visualization
AJ Christensen Introduction to NASA and Early Career Pathways
Matt Bruzek Computational Methods for Evaluating Transonic Aircraft Buffet
Daniel Maldonado Supercomputing in Supersonics: Quiet Supersonic Flight
Marian Nemec Becoming a NASA Intern and Exploring the World of CFD Simulations
Patricia Ventura Diaz How NASA Manages User Access Using Arbiter2
Dave Chau Managing and Patching Multiple Architectures at NASA
Don Story Designing Ansible for Large Scale Environments
Alex Pelkey