Scientists Develop AI Model Using Earthquake Data

The new "SeisModal" AI foundation model aims to explore big questions about science.

Published on Feb. 12, 2026

Got story updates? Submit your updates here. ›

Researchers from five U.S. Department of Energy national laboratories have developed a new AI foundation model called "SeisModal" that uses data from the world's largest repository of earthquake information. The goal is to create a broad, trustworthy AI tool that can be adapted to tackle various scientific questions, with a focus on nuclear nonproliferation research.

Why it matters

Foundation models are crucial tools for AI research, but few have been built with a specific focus on scientific applications. SeisModal aims to provide a robust, multimodal AI foundation model that can analyze diverse data types like seismic waveforms to support a wide range of scientific inquiries, particularly in the area of nuclear nonproliferation.

The details

The SeisModal model was developed through the Steel Thread project, which is funded by the National Nuclear Security Administration. It integrates data on earthquake intensity, location, timing, waveforms, text, and imagery to create a comprehensive picture of seismic events. The model's ability to reason over complex time-series data like seismic waveforms sets it apart from many current large language models. The goal is to build a trustworthy AI tool rooted in transparent, high-quality scientific data that can be adapted to explore various scientific concepts.

  • The Steel Thread project was discussed in an invited talk at the annual Joint Statistical Meetings in Nashville last summer.

The players

Karl Pazdernik

A chief data scientist at Pacific Northwest National Laboratory and the science lead of the Steel Thread team.

Sai Munikoti

A PNNL scientist and the lead architect of the SeisModal model.

Ian Stewart

A PNNL scientist and the lead architect of the SeisModal model.

National Earthquake Information Center

The organization that maintains the dataset of over 16,000 seismic events used to train the SeisModal model.

National Nuclear Security Administration

The agency that is funding the Steel Thread project to develop the SeisModal AI foundation model.

Got photos? Submit your photos here. ›

What they’re saying

“We're creating a foundation model with broad capability that can be applied to multiple problems in science with minimal retraining for each application.”

— Karl Pazdernik, Chief data scientist, Pacific Northwest National Laboratory

“Creating an AI foundation model whose goal is to understand scientific concepts can be a big lift, but it can have many applications beyond seismology.”

— Karl Pazdernik, Chief data scientist, Pacific Northwest National Laboratory

“Since we want our models to be rooted in science, a major focus of our project is also to make sure that any model that we build is trustworthy. To evaluate its trustworthiness, we need to understand the training data, be sure of its origin, and describe the security and usability of the model.”

— Karl Pazdernik, Chief data scientist, Pacific Northwest National Laboratory

“SeisModal can reason over complex time series data such as seismic waveforms, which is an advance over many current large language models. The ability to detect these signals and other uncommon data types opens the door to a wider variety of scientific analysis methods that were previously unavailable.”

— Ian Stewart, PNNL scientist

What’s next

The Steel Thread team plans to continue developing and refining the SeisModal AI model to make it an even more powerful and versatile tool for scientific research, with a focus on nuclear nonproliferation applications.

The takeaway

The creation of the SeisModal AI foundation model demonstrates how cutting-edge AI technologies can be leveraged to tackle complex scientific challenges, particularly in sensitive areas like nuclear nonproliferation. By building a trustworthy, multimodal AI model rooted in high-quality data, researchers are paving the way for new breakthroughs in a wide range of scientific fields.