Atomic Canyon is a six month old company that is developing AI tools to improve the efficiency of routine tasks associated with developing, licensing, building, owning and operating nuclear plants. Their first product, called Neutron, uses AI to modernize searching the Nuclear Regulatory Commission’s 52 million page collection of publicly available documents that are currently accessible through the somewhat cumbersome Agencywide Documents Access and Management System (ADAMS).
Trey Lauderdale, Atomic Canyon’s founder, spent the first 15 years of his career in the digital medicine field. At an inflection point in his career, with the freedom to live anywhere, he created a decision matrix to help him and his wife choose a place to live and raise their two young sons. San Luis Obispo, CA earned the highest score, with an excellent public education system as one of the contributing factors.
After finding their home and moving towards closing the purchase, Trey and his wife learned via real estate disclosure documents that they would be living within 10 miles of the Diablo Canyon Nuclear Power Plant. As members of a generation who learned most of what they new about nuclear energy from The Simpsons, they were initially leery.
But they quickly realized that the plant’s skilled, dedicated and well compensated employees and its property tax payments were major reasons that the schools and other aspects of the community had earned such high scores on the “place to raise our children” decision matrix.
After becoming a member of the community and conversing with local nuclear professionals, Trey decided to learn as much as he could about nuclear energy and the nuclear industry. He recognized that he and his skilled colleagues could build tools that could address obstacles that slowed work and added costs.
Atomic Canyon has just announced a cooperative project with the Oak Ridge National Laboratory (ORNL) that will train ORNL’s Frontier – currently the world’s fastest supercomputer, capable of more than a quintillion calculations per second – how to understand nuclear terminology. The resulting model will not be trained on proprietary or safety related information on the design and operation of nuclear power, but it will help analyzing the deep library of regulatory guides, inspection reports, and other publicly available documents to assist in increasing safety and accountability.
The products (models) created by the partnership will be open source and available to become part of the toolbox for other developers.
Trey and I had a fascinating conversation. I think you will agree.
Left to right in photo: Trey Lauderdale, Atomic Canyon CEO Kristian Kielhofner, Atomic Canyon CTO Richard Klafter, Atomic Canyon Lead AI Architect Tom Evans, ORNL Research Scientist Photo Credit: Genevieve Martin, Oak Ridge National Laboratory
Knox News provides a local perspective on Atomic Canyon’s project using Frontier: AI for nuclear plants? ORNL supercomputer’s new task is no sci-fi – it’s a clean energy win