Tech companies are betting that nuclear power can help deliver the electricity needed for AI advancements. Yet, nuclear’s slow pace might hinder immediate demands. Entrepreneur Trey Lauderdale believes AI can speed up the nuclear industry.
Nuclear power and artificial intelligence: an unlikely duo at first glance, yet they may soon intertwine as tech companies aim to harness nuclear energy to meet the insatiable demand for electricity, especially in the realm of AI. As data centers spring up to support AI operations, their dependence on fast and reliable energy sources grows urgent. However, the nuclear sector, often criticized for its slow development processes, presents challenges in meeting this immediate energy need.
Enter Trey Lauderdale, a serial healthcare entrepreneur with a newfound passion for nuclear energy. His journey began in his hometown of San Luis Obispo, California, where interactions with professionals from the Diablo Canyon Power Plant opened his eyes to the struggles of the nuclear industry. “They’re like the coaches of our flag football team,” he fondly recalls. Through casual conversations, he discovered an overwhelming amount of documentation—around 2 billion pages concerning the operations of the plant—which sparked an idea: could AI streamline this monumental paper problem?
With this vision, Lauderdale founded Atomic Canyon, a startup dedicated to leveraging AI to assist nuclear engineers and technicians in locating pertinent documents efficiently. Initially funded from his own pocket, the venture found its footing in late 2024 when it secured a significant contract with Diablo Canyon. This partnership not only validated Lauderdale’s concept but also attracted interest from other nuclear companies looking to embrace modern solutions.
Atomic Canyon soon closed a $7 million seed funding round led by Energy Impact Partners, with contributions from various investors, including Commonweal Ventures and Plug and Play Ventures. This financial backing was crucial for scaling the technology and expanding their operations to meet the needs of the nuclear sector.
In the early stages, Lauderdale and his team faced challenges when testing different AI models, realizing that the AI often struggled with nuclear terminology. “We quickly realized the AI hallucinates when it sees these nuclear words,” he remarked, highlighting the unique complexities of the industry. To overcome this, Lauderdale sought collaboration with Oak Ridge National Laboratory, a premier research facility equipped with one of the fastest supercomputers globally. This partnership was fruitful, as Atomic Canyon was granted access to 20,000 GPU hours for further AI development.
Atomic Canyon’s innovative approach includes leveraging sentence embedding technology for indexing nuclear documents. Their method employs retrieval-augmented generation (RAG), which allows AI to formulate responses based on specific documents, reducing the likelihood of hallucinations. This meticulous approach aims to improve searchability within the vast archives of nuclear documentation—an essential first step in streamlining processes and ensuring compliance.
Presently, Atomic Canyon focuses on optimizing document search functions, as the stakes are lower compared to other applications of AI. Lauderdale expresses a cautious optimism: “One of the reasons we’re starting with search is that getting it wrong may lead to minor frustration but not risks to plant safety.” By initially concentrating on creating accurate search results, the team can lay a solid foundation for more advanced uses of AI in document creation later on.
Looking toward the future, Lauderdale envisions a scenario where Atomic Canyon’s AI could draft preliminary documents enriched with references and critical information. He emphasizes that while AI can draft content, human oversight will always remain integral to the process. “You are always going to have a human in the loop here,” he states, highlighting the balance between technology and human expertise essential in the nuclear field.
While Lauderdale has not mapped out a specific timeline for expanding beyond search capabilities, he underscores the importance of mastering this foundational aspect first. With an enormous corpus of existing documents, the potential for improving search functionalities presents a lengthy runway of opportunity for Atomic Canyon. Each enhancement brings them one step closer to revolutionizing the nuclear industry’s document management through AI, potentially transforming the energy landscape as a result.
Inspired by: Source
