Surya Science:
A hands-on workshop
Join us to learn how to use Surya: A foundation model for SDO data in your research
Overview
While the live workshop is complete, the Surya Science initiative is ongoing. We invite stakeholders and researchers who were unable to attend to explore collaboration opportunities.
Surya is a foundation model specifically developed by NASA and IBM for heliophysics, trained on 9 years of high-resolution observations from the Solar Dynamics Observatory (SDO). The release of this domain-specific foundation model marks a pivotal moment in how we study the Sun, rapidly accelerating what is possible. The workshop explored how Surya can be adapted to support a range of downstream scientific tasks, with a hands-on focus building on the ideas, opportunities, and research directions this capability enables in actual workflows.
For further technical documentation, insights, or to discuss future applications, please reach out to our team at suryascience@trillium.tech
-
9:00 - 9:30: Welcome to Surya Science: A Hands-on Workshop:
Welcome from the organizers, with briefing on how the workshop will work.
9:30 - 11:30: Introduction to Surya-FM: (General registrants included)
Foundation models for science
Johannes Schmude | IBM
Surya 101
Sujit Roy | NASA-IMPACT
What Surya can do for heliophysics?
Andrés Muñoz-Jaramillo | Southwest Research Institute, Colorado, USA
Introduction to the SuryaBench dataset
Dinesha Vasanta Hegde | University of Alabama in Hunstville
11:30 - 13:00: Lunch break13:00 - 13:25: Downstream science demos (General registrants included)
A set of demos from the science team showcasing a selection of downstream science use cases
13:25 - 13:45: HelioAI - what’s next? (General registrants included)
Mike Heyns | Trillium Technologies
Shing Fung | NASA
13:45 - 16:15: Dataloaders deep dive
Tutorial session on best practices for the hands-on aspect of the workshop
Explanation of PyTorch datasets and dataloaders
Working session to implement dataloaders
16:15 - 17:00: Establishing baselines
Explanation of baselines
Daily wrap up
Async work on baselines
-
9:00 - 10:00: Share Back Session
Morning briefing session, with share back session on learnings from work on baselines
10:00 - 16:00: Finetuning and setting up experiments
Explanation of Surya-FM architecture and how to modify it
Tutorial on logging experiments
Explanation on putting together an end-to-end application for a downstream use case
Working session on finetuning applications
16:00 - 16:45: Guest Speakers: Industry Panel (General registrants included)
Panel session on the role of foundation models from an industry perspective
Dr. Juan Bernabe Moreno | IBM Europe
Dan Crichton | NASA/JPL
Massy Mascaro | Google Cloud
16:45 - 17:00: Daily wrap up
-
9:00 - 10:00: Share Back Session
Morning briefing session, with share back session on learnings from work on finetuning
10:00 - 15:00: Validation and evaluation
Explanation of evaluation set, metrics, and how to compare models
Working session on validation approach for downstream use cases
Continue experiments with agreed validation approach
15:00 - 16:00: Presentation preparation and final experiment design
Facilitated session on presentation preparation using workshop templates and final experiment design
16:00 - 16:45: Guest Speakers: Agency Panel (General registrants included)
Panel session on agency perspectives, priorities and future aspirations
Dr. Madhulika Guhathakurta | NASA Heliophysics
Rahul Ramachandran | NASA-IMPACT
Therese Moretto Jorgensen | NASA HQ
Robert Redmon | Director of NOAA Center for AI
16:45 - 17:00: Daily wrap up
-
9:00 - 10:00: Share Back Session
Morning briefing session, with share back session on learnings from work on validation
10:00 - 11:30: Presentation finalization
Finalize presentation, with drop-in sessions available
13:00 - 16:00: Lightning presentations and next steps (General registrants included)
Final lightning presentations showcasing outputs from the week, with dedicated session to map out next steps and start populating a publication template
16:00 - 16:45: Guest Speakers: Science Panel (General registrants included)
Panel session around applying foundation models in the wild and how do we make decisions using foundation models?
Raul Ramos-Pollan | Universidad de Antioquia
David Fouhey | NYU
Tzu-Wei Fang | NOAA Space Weather Prediction Center
16:45 - 17:00: Workshop closing
-
9:00 - 9:25: Welcome to FDL Day
Guest Speaker: Dr Madhulika Guhathakurta
9:25 - 10:55: Heliolab: Technical Showcase
Presentations from Heliolab 2025 research teams:
- ARCADE
- Decoding the Solar Wind
- Multimodal Flare Prediction
- Ionosphere-Thermosphere Twin
- Orchestrator11:10 - 11:50: FDL: ‘Big Think’ Introduction
FDL 'Big Think' briefing session and initial ideas pitches, following by start of ideation session
11:50 - 12:30: Guest Speaker: Google AI co-scientist
Dustin Sell (Google)12:30 - 13:30: Working Lunch
13:30 - 13:45: Guest Speaker: Google Space
Sammo Morgan (Google)13:45 - 14:25: FDL: ‘Big Think’
Facilitated FDL 'Big Think' ideation session
14:25 - 14:40: Break
14:40 - 15:20: FDL Big Think
15:20 - 16:20: Lightning pitches
Lightning pitches of 'Big Think' ideas
16:20 - 16:30: Wrap up and next steps
About Surya
Surya is a 360-million-parameter Heliophysics foundation model trained on more than 200 TB of solar data collected over nine years by the Solar Dynamics Observatory. Surya leverages advances in AI to analyze vast solar datasets, improving our understanding of solar eruptions and enabling more accurate space weather forecasts that protect satellites, power grids, and communication systems. Surya was developed by NASA in partnership with IBM and a team of researchers from multiple institutions.
In this workshop you will learn all about Surya and get the opportunity to develop downstream applications that use this powerful model.
Watch this video to learn more about Surya
Read more about Surya
Partners
