Agencies across government are under pressure to build artificial intelligence skills into their workforces. The Marine Corps has picked a specific mechanism for doing it — and the design detail that matters most, according to the program's lead at the Naval Postgraduate School, is that every fellow arrives with a real problem to solve.
The USMC-NPS AI Fellowship launched as a pilot flowing from the service's 2024 AI Strategy, which set out to accelerate AI research and build AI talent inside the Corps. It was modelled on the "Phantom Program," an Air Force initiative run with MIT. Two slots went to Marines under that arrangement, and service leadership wanted a way to involve more people.
A command-endorsed use case is the entry ticket
"It's a commitment for the Marine to dedicate their full-time primary effort to the fellowship for five months," said Christopher Paul, the U.S. Marine Corps chair for information at NPS, speaking at Federal News Network's Cyber Leaders Exchange 2025.
Admission requires a colonel-level endorsement letter releasing the Marine for the duration — and, critically, a command-approved use case.
"There's some problem at that command, some challenge that — at least on its face — appears amenable to a solution or improvement with data and perhaps with AI or machine learning that the command endorses," Paul said. "So yes, it's a chance for us to take some Marines and do a little bit of upskilling. But there's also this problem focus, so that by the end of the five months, they've got, hopefully, at least a prototype solution to take back to their command and change how they do business, or improve decision-making or make processes more efficient."
The first cohort is five Marines working four use cases, ranging from analysing imagery in preparation for amphibious landings to streamlining the paperwork around maintenance activities. Ranks run from captain to corporal, plus one Marine Corps civilian. Other projects across the programme have included using large language models to cut maintenance documentation burden and automating cyber vulnerability testing.
Three bins of AI aptitude
The fellowship sits alongside an NPS AI task force considering AI education across the military services more broadly. Paul said educators there have started sorting the need for AI aptitude into three bins depending on a service member's role: enough understanding to know how and when AI can be used; enough expertise to actually train and modify models; and enough familiarity to follow the terminology.
"We're thinking that this fellowship is aimed toward folks who are either in or on their way toward that middle bin, that one of actual AI expertise," he said. "They're going to come here, they're going to solve a problem, but in learning that, they're going to learn about how to solve other problems — and they're going to see as they survey existing courses, 'OK, I don't need to take that course online. I do need to take this course online.' … And sometimes, when you're working on a real problem, the experience is different than what's in the textbook or what's in the online video."
How the five months are structured
Despite the five-month duration, the first group spends only about three weeks in residence at the NPS campus in Monterey, California. The remainder is spent on their individual use cases, punctuated by a midpoint progress report back at NPS and a final presentation at the close of the fellowship.
NPS and the Marine Corps are already preparing a second cycle, and Paul said officials have discussed opening the programme to Navy service members.
What success would unlock
Paul set a deliberately modest bar for proof of concept. "If this is modestly successful — if at least one of these four projects is solidly successful — then that is a proof of concept that demonstrates that, hey, with a relatively modest contribution of manpower, a real problem can really be at least partially solved, so that means we should do more of this … and maybe we can connect this more formally with the Marine Corps digital transformation teams."
Beyond that, he sees an institutional opening. "The Air Force has their AI accelerator at MIT. The Army has an AI accelerator at Carnegie Mellon. Perhaps the naval services need an AI accelerator. Perhaps that could be here at NPS, and some success from this fellowship could recommend NPS as a good home for that."
A stronger showing could push the model outward rather than upward — toward mobile training teams seeding digital transformation units across the fleet. "Because with a little bit of faculty advice and a lot of independent learning, any smart Marine can learn to do these things," Paul said.
Your membership also unlocks:
