Deep Learning Breakthrough at DAF-MIT AI Accelerator Modernizes U.S. Nuclear Surveillance Mission

  • Published
  • By DAF AIA Public Affairs
  • DAF AIA
CAMBRIDGE, Mass. --  A blend of military expertise and cutting-edge academic research has produced the Department of the Air Force – Massachusetts Institute of Technology Artificial Intelligence Accelerator’s latest national security advancement. This success stems from Tech. Sgt. Brian J. Wright, one of the latest graduates of the AI Accelerator’s (AIA) Phantom Fellowship Program, who developed a breakthrough deep learning solution that transforms a vital data analysis process for the Air Force.

These fellows, known as Phantoms, are selected through a highly rigorous and competitive selection process and come from diverse career backgrounds. Phantoms embed themselves into AIA research projects, contributing their specialized knowledge and expertise to a range of projects, while also innovating and developing novel processes to facilitate the successful integration of AI research into practical applications. Simultaneously, each Phantom undertakes an individual impact project, leading and translating research into actionable solutions for their home units.

“The Phantom Fellowship is truly a remarkable program that allows the AIA to tap into talent across the DAF and couple these Airmen, Guardians, and civilians with researchers to pursue tactical AI applications,” said Master Sgt Chasen Milner, Phantom Program Manager at the DAF-MIT AI Accelerator. “Moreover, it equips Phantoms to return to operations ready to assess, field, and manage AI technologies.”

Wright, a Cohort 11 Phantom, is a highly skilled senior data scientist with more than a decade of analytics experience in the Air Force. He holds an undergraduate degree in Mathematics and is pursuing a Master of Science in Data Analytics from the Georgia Institute of Technology. Bringing this robust technical background and deep operational knowledge into the fellowship, he tackled the challenge head-on to advance the missions at his home unit, the Air Force Technical Applications Center (AFTAC), headquartered at Patrick Space Force Base, FL.

AFTAC plays a critical role in countering next-generation nuclear threats while ensuring compliance with international nuclear treaties. Consisting of more than 3,600 sensors worldwide, AFTAC operates and maintains the largest sensor network in the U.S. Air Force.

As AFTAC’s global sensor coverage expands and data volume continues to grow, so does the need for faster, more scalable analysis methods. In today’s rapidly evolving threat environment, delivering timely and accurate nuclear assessments isn’t just a technical challenge—it’s a national security imperative.
To meet that challenge, AFTAC is reimagining the center’s ability to detect, verify, and analyze nuclear activity in real time, adapting its capabilities to keep pace with the speed, complexity, and scale of emerging threats.

AFTAC’s efforts include analyzing radioactive isotopes using a highly precise process called beta-gamma coincident spectrometry. This technique allows experts to identify unique features of nuclear materials. “Current analysis methods require full human involvement at every step,” Wright explained. “To keep up long-term, we would have to grow our team at the same rate as the data, and that’s simply not a sustainable solution.”

This realization drove him to carry the problem to the DAF-MIT AI Accelerator, where he dedicated his fellowship to crafting a solution through applied research. “My mentor, Janelle Ghanem, a Machine Learning Engineer at Lincoln Laboratory, kept me motivated and on track through monthly check-ins. She also reviewed my code and slides to make sure my final presentation was accurate and error-free,” Wright said. The result was his development of several deep learning neural networks aimed at automating and optimizing AFTAC’s data analysis processes.

These new tools are capable of performing the same amount of work that would take an entire team a full week in just under six seconds while still hitting remarkable levels of accuracy. Most impressive of these are convolutional neural networks, a method originally invented for image processing, which drastically enhances mission resiliency against data degradation, a common challenge in AFTAC’s operational environment.

“While they are not without limitations,” noted Wright, “the levels of performance we have achieved give us confidence that in the future we will be able to delegate a significant portion of non-critical analysis to these machine learning networks. This ensures that human expertise remains focused on the most crucial data, a classic ‘human-in-the-loop’ system that optimizes both speed and expert oversight.”

The impact of these innovations was clear when Wright presented his research outcomes to AFTAC Commander, Col Creighton Mullins.
“What TSgt Wright accomplished during his Phantom Fellowship is transformational,” said Col Creighton Mullins, AFTAC commander. “By harnessing the power of artificial intelligence, he’s delivered a scalable solution that accelerates how we process and analyze critical data. That kind of speed and efficiency is essential in today’s time-sensitive threat environment.”

“AFTAC is not only a nuclear surveillance center but also a hub of innovation,” Mullins added. “TSgt Wright exemplifies the warrior ethos: he leveraged his technical skills to strengthen our ability to monitor and respond to nuclear threats across the entire spectrum, from development to deployment.”
Recognizing the value of TSgt Wright’s research, Col Mullins directed that Wright’s AI-driven solutions be operationalized, strengthening AFTAC’s ability to meet the needs of the warfighter and national decision makers.

“The work of Phantoms like Tech. Sgt. Wright showcases the value of continued investment in fundamental research to impact operations,” said Col Scott Ruppel, Director DAF-MIT AIA. “By leveraging our collaboration with MIT and applying innovative thinking, we are modernizing our capabilities to face the complex challenges of our time.”

Through the Phantom Program, the DAF-MIT AI Accelerator has once again demonstrated its value as a catalyst for innovation, proving that with the right tools, even the most formidable challenges can be overcome. Wright’s journey from problem identification to impactful solution is a testament to the power of collaboration between military and academic experts—a collaboration that promises to keep the Department of the Air Force on the cutting edge of technological advancement.