Efficiency first: Sandia’s new director balances AI drive with deterrent work

Laura McGill, two weeks into her tenure as Sandia National Laboratories director, addressed the organization’s New Mexico staff on May 14. She stated that the lab would double down on its core role as the nation’s nuclear-weapons system integrator while also scaling up digital-engineering and AI projects. She announced a projection of up to $5 billion in construction-related spending over the next 10 years, mostly in New Mexico, for major and minor capital projects to support Sandia’s mission. “This is a win-win situation,” she said. “We will be furthering our mission and providing a stable source of work and economic opportunity for design firms, construction companies and the skilled trades people across our state in this region.”

McGill kept the focus squarely on the stockpile. As lab director, she will work to help”ensure that all the advanced components and technologies that are in the weapon systems, and the delivery systems, work together flawlessly, from the initial design and all the way through the retirement of the system and dismantlement,” she said. The functional criterion, she said, is binary: A nuclear weapon should work “when authorized by the President of the United States and never under any other circumstances.”

In her remarks, McGill also emphasized Sandia’s focus on artificial intelligence, framing it as a transformative technology with acute urgency for national security and an efficiency driver. “Sandia is really leaning into AI,” she said. “AI allows us to make better decisions, deliver better solutions for the nation, and be more efficient stewards of taxpayer dollars.”

She also noted: “As our Energy Secretary Chris Wright said, this is a ‘Manhattan Project moment’ for us in terms of the urgency of bringing AI into the national security space. And he’s right.” McGill highlighted that Sandia was “the first facility to offer an isolated, internal generative AI tool, Sandia AI Chat, that our employees have been using for quite some time now with access to curated lab information.” While acknowledging that “there are considerations, there’s risks and constraints with AI,” she asserted, “they are actually manageable, and we don’t shy away from the complexity. We engineer the solutions.”

McGill pointed to two key AI efforts. She highlighted “Sandia AI Chat,” calling it “an isolated, internal generative-AI tool…that our employees have been using for quite some time now with access to curated lab information.” According to Sandia’s LabNews, this system is the Labs’ own instance of a large language model (based on ChatGPT technology but run in Sandia’s own Azure Cloud “box”), operating within a secure, Sandia-controlled environment. This ensures that employee queries and proprietary lab data remain confidential, are not shared with external entities like Microsoft or OpenAI, and are not used to train public AI models.

The second major AI effort McGill described involves applying “high-power computing” and AI algorithms “to help point us in the direction of new materials that maybe haven’t been invented yet: new alloys and metals that have specific characteristics that we need to support the nuclear deterrent, or other programs.”

McGill also noted that quantum science is “obviously really important” to the lab. 

This conference came just days after a May 12 announcement that Sandia helped save the Department of Energy $439 million through the Supply Chain Management Center program, more than any other DOE site.