SoarTech was one of 10 small businesses awarded more than $11 million, collectively, through the U.S. Army’s Applied Small Business Innovation Research Program, to develop artificial intelligence and machine learning (ML) solutions with the potential to address ground warfare challenges for the U.S. Soldier.
“Our DeepSynth project will explore an innovative approach to synthetic data creation to tackle the data bottleneck in training AI/ML systems for mission critical applications,” said Mike van Lent, CEO. “The Army’s SBIR program is a critical enabler of the innovative small businesses that keep the Army strong and soldiers safe.”
The awards are in response to three contract opportunities released by the Army Applied Small Business Innovation Research Program for U.S.-based small businesses to propose technology in the areas of electronic warfare, sensors, information systems technology, battlespace environments and human systems.
SoarTech, working with Leidos and Attollo, will develop DeepSynth, a tool to quickly generate synthetic Electro Optical (EO) and Synthetic Aperture Radar (SAR) training data based on input for analysis.
“DeepSynth will combine technology previously developed across multiple projects by SoarTech and Leidos into a single tool,” explained Chris Kawatsu, SoarTech’s principal investigator. “SoarTech will bring our extensive experience in deep learning for military applications and our state-of-the-art capabilities in generating synthetic EO data.”
Leidos provides advanced capabilities in synthetic SAR data and their application to training deep networks, while Attollo provides subject matter experts and labeling expertise for small datasets.
“DeepSynth will allow the U.S. and its allied forces to rapidly create synthetic data for training and evaluating models,” said Kawatsu. “This results in a large reduction in cost compared to creating datasets from real data, which can require decades of worker years to produce.”
“The U.S. and its allied forces use AI to make decisions faster on the battlefield, providing a tactical advantage over adversaries — but the data to train that artificial intelligence is a critical input,” said Bharat Patel, technology and industry liaison officer with Army PEO Intelligence and Electronic Ware & Sensors and technical point of contact for one of the SBIR topics. “So, being able to augment that training data with synthetic data to train and evaluate models can speed up the development life cycle with significantly less sustained cost.”
In addition to benefiting the Army, the Army Applied SBIR Program offers a valuable opportunity for small businesses to interact with Soldiers and technical and operational subject matter experts, who provide insight into technology needs and guide small businesses through the Army research and development ecosystem. Small businesses are teamed up with technical points of contact who serve as a resource for companies as they mature their technologies for insertion into Army programs of record.