You’ve heard of nuclear plants powering the data needs of AI. Now, get ready for AI to help the nuclear sector in return.
Atomic Canyon is an AI startup working to help nuclear energy companies navigate the vast labyrinth of regulatory requirements. The startup announced yesterday that it’s partnered with Oak Ridge National Laboratory on a large-scale project parsing nuclear-related language.
“We think this is such an important foundational element to enable artificial intelligence across the entire nuclear power space,” Trey Lauderdale, CEO of Atomic Canyon, told Ignition. “We want to put it out there, and want to be open and transparent so everyone can leverage it.”
The partnership details: Atomic Canyon will work with Oak Ridge to develop a sentence-embedding model, which will help its AI to understand nuclear terminology in context.
- Atomic Canyon will gain access to the Frontier supercomputer—the world’s fastest supercomputer—to train its AI to better understand the complex vocabulary of the nuclear sector.
- The project was approved through Oak Ridge Leadership Computing’s Director’s Discretion, meaning that Atomic Canyon won’t have to pay for the privilege of using the Frontier supercomputer.
- Initial results of the project are expected on a super-fast timeline—by this summer, per Lauderdale.
The company plans to make all its results open source, helping other nuclear AI players train their models, too.
“We want hundreds of AI companies coming into this space,” Lauderdale said. “I think there is endless amounts of work that’s going to need to be done to help make the nuclear power industry more effective, more efficient, and this sentence embedding model helps make the artificial intelligence more accurate and more reliable when it comes to nuclear terminology.”
Back to basics: Atomic Canyon hit the scene this year with Neutron, an AI product that helps nuclear companies search through the NRC’s vast database of documentation.
- The company’s gotten a ton of interest from the nuclear sector. It currently has ~50 regular users on Neutron and plans to onboard more.
- The team is small but mighty, with ~10 full- and part-time software engineers.
Right now, Atomic Canyon is working on improving Neutron through a number of pathways, including this partnership with Oak Ridge—and it’s making the results publicly available. Lauderdale said that’s because the revenue opportunities for the space lie beyond these foundational AI steps.
“It’s not just the applications themselves,” Lauderdale said. “It’s the deployment, the implementation, the integration to other systems.”
Lead Reporter of Ignition
