California-based startup Atomic Canyon has formed a strategic partnership with the US Department of Energy’s (DOE’s) Idaho National Laboratory (INL) to further develop AI (artificial intelligence) for the nuclear industry. The partnership will establish critical standards for AI adoption in nuclear facility operations, addressing an industry need for objective evaluation methods for generative AI systems in nuclear environments. The aim is to publicly release the first comprehensive benchmark suite for evaluating retrieval-augmented generation (RAG) and large language models (LLMs) in nuclear applications.
Nuclear facilities and national laboratories store decades of operating logs, maintenance records, and engineering drawings – much of it in scanned PDFs or proprietary repositories. The new benchmark suite will enable nuclear operators to evaluate AI systems that can access public nuclear documentation through retrieval-augmented generation techniques.
The benchmark suite aims to address a fundamental challenge in nuclear AI deployment: evaluating systems that must work with both public nuclear documentation and facility-specific technical data. Existing AI evaluation benchmarks are insufficient for the nuclear sector’s unique requirements for safety, regulatory compliance, and security.
Over six months, the project will curate datasets, define benchmark tasks and evaluation metrics, and produce comprehensive documentation, including a technical summary, for community distribution via open-source repositories. All data, definitions, and criteria will be released under open-source public licenses, ensuring the nuclear and AI communities can benefit from this foundational work. Atomic Canyon will develop and release the accompanying evaluation software and tools under permissive open-source licences.
“This partnership represents a pivotal moment for the nuclear industry’s digital transformation,” said Atomic Canyon CEO Trey Lauderdale. “Nuclear facilities require the highest standards of safety and compliance, and our collaboration with INL will establish the gold standard for evaluating AI systems in nuclear environments. These benchmarks will give utilities, regulators, and technology vendors a common framework for ensuring AI systems meet the rigorous demands of nuclear operations while maintaining the absolute highest levels of safety and security.”
INL will contribute its unique nuclear subject-matter expertise and access to curated public datasets. The benchmark tasks will focus on real-world nuclear workflows, including document retrieval, regulatory compliance checks, and answering technical questions.
Christopher Ritter, Division Director of Scientific Computing & AI and Director of INL’s Digital Innovation Center of Excellence (DICE) said: “This partnership with Atomic Canyon will establish the benchmarks needed to effectively deploy AI across the nuclear enterprise. By creating standardized evaluation methods, we’re enabling the industry to confidently adopt AI technologies that can accelerate nuclear energy deployment for America.”
Atomic Canyon was founded in 2023, and began by cataloguing documents at the Diablo Canyon NPP. It sorted through reems of papers as the NPP prepared for relicensing after the state abandoned its plan to permanently phase out fission energy. The company, which emerged from stealth in March 2024, develops AI-powered search and generative AI tools specifically for the nuclear power industry. The company’s flagship product, Neutron Enterprise, provides AI-powered capabilities across vast repositories of technical documentation at nuclear facilities. By securely connecting to internal data sources and external authoritative technical and regulatory data sources, Neutron Enterprise enables nuclear professionals to access critical information in seconds rather than hours.
In May 2024 Atomic Canyon launched a project with the US DOE’s Oak Ridge National Laboratory (ORNL), to build a safe, efficient open-source AI model for the nuclear sector. Atomic Canyon is using ORNL’s Frontier supercomputer – the world’s fastest supercomputer – to train the company’s AI model to understand complex nuclear terminology. Atomic Canyon is training its AI-powered search tool on Frontier after being awarded time on the computer by DOE. Atomic Canyon recently raised $7m in seed funding led by Energy Impact Partners to accelerate deployments across the nuclear industry.
The company is now in discussions with utilities representing “40% to 50% of the US nuclear fleet,” Lauderdale told Latitude Media. “A lot of these utilities, their IT teams came and gave everyone broad access to AI tools. There was a period of excitement where everyone said, ‘This is great; look at this cool toolset’, but then people started to try to do more advanced functions within their selected technology vendors’ ecosystems and realised the limitations hit way faster than expected.”
He explained: “It takes a lot of effort and research and development to create and build applications that will actually work reliably and at scale.” After visiting several nuclear plants and talking to them about their software needs it became clear that the first area of focus was just “helping them find documents effectively”.
He added: “We believe it’s our responsibility as the lead in generative AI and nuclear to create the first benchmark for the industry in how these large language models fare when they’re used. To do that, we wanted to make sure we had a neutral third party to join us. That’s where Idaho National Labor came in.”
