An M777 155mm howitzer fires at Yuma Proving Ground's Kofa Firing Range,
America's longest instrumented overland artillery range. Data analysis of acoustics sensors for munitions testing at YPG was the topic of the 2023 AI Challenge.

Body armor plate used in the 2024 AI challenge that sought to improve analysis of flaws in Soldier body armor plates.

ABERDEEN PROVING GROUND, Md. – The U.S. Army Test and Evaluation Command has announced the focus of its third annual AI Challenge, which kicks off June 4.

The objective is to develop a Large Language Model (LLM)-based tool to address any existing job-related problem within ATEC.

LLMs are tools capable of understanding and producing human-like text. They are able to assist with tasks such as drafting documents, answering questions and summarizing information.

Participants must submit their proposals by July 9, and the challenge ends on Aug. 15. A panel of judges will evaluate the solutions based on their effectiveness, innovation and return on investment. The winners will be announced at the 2025 ATEC Data Summit taking place Sept. 16-18.

ATEC started its annual AI Challenge in 2023 to empower its workforce to develop cutting-edge AI skills while solving real-world ATEC challenges using AI-ready data sets.

Last year’s challenge focused on how to improve the analysis of flaws in Soldier body armor plates. Prior to the challenge, analysts would perform the flaw analysis by examining X-ray images, which was a time consuming process.

The winning team in 2024 utilized a cutting-edge model for image analysis as the solution. This model used, known as the Vision Transformer (ViT), processes images by treating them as a sequence of patches. Using the ViT sped up the flaw analysis process and successfully gave the analyst time back to investigate the root cause of any identified flaws.

In 2023, participants provided solutions to improve data analysis of acoustical sensors for munitions testing at Yuma Proving Ground. The winning solution is currently in use at YPG, saving the Army hundreds of man hours of work.

Unlike last year’s competition, which was open to the acquisition community across the Department of Defense, this year’s challenge is only open to ATEC military or civilian personnel, with approval obtained from their chain of command.

Maj. David Niblick, military evaluator from the Analytics, Artificial Intelligence and Digital Engineering Directorate within the U.S. Army Evaluation Center, an ATEC subordinate organization, expressed his excitement about this year’s competition.

"I'm ecstatic to see ATEC lean into these exciting LLM technologies,” said Niblick. “This challenge is one way that ATEC will gain efficiency and efficacy while learning necessary skills in the modern age."