Synthetaic, which has developed a field-proven artificial intelligence tool that rapidly searches for images in datasets without labeling, on Tuesday said it has raised $15 million in a new funding round to further grow and expand its presence in national security and commercial markets.
The Series B funding round was led co-led by Lupa Systems
and TitletownTech, with participation from IBM’s [IBM] venture arm and Booz Allen Hamilton [BAH]. In 2022, Booz Allen also participated in a $13 million funding round for Synthetaic that was led by Lupa Systems (Defense Daily, March 23, 2022.)
Defense and intelligence customers currently account for more than half of Synthetaic’s sales but “our perfect balance” would be a 50-50 split between the national security and commercial markets, Corey Jaskolski, founder and CEO, told Defense Daily. “Our goal is to stay dual use,” which he said will help balance the ups and downs in the respective end markets.
Synthetaic, which is based near Milwaukee, has nearly 50 employees, including one in Washington, D.C., where the company is actively hiring, he said.
Jaskolski said that the federal market was not on his radar when he founded Synthetaic but the company’s technology is a “great fit” for defense and intelligence agencies that need answers today or tomorrow, not in the eight months it typically takes to create an AI project.
Synthetaic’s AI tool is called RAIC, which stands for Rapid Automatic Image Categorization. Unlike traditional AI models that that are labor intensive due to the need to manually label images in datasets, RAIC just needs a single image, which could be a drawing or an overhead image something like a satellite image to use to query a dataset, Jaskolski said.
RAIC works like generative AI, allowing a user to search a dataset for whatever it is they are looking for, he said.
When a Chinese high-altitude balloon was spotted over the U.S., RAIC was used to query satellite imagery produced by Planet Labs [PL] satellites to track where the ballon had been. Jaskolski said when the software was used to search back in time through archives, it instantly found Chinese balloons that had flown over Columbia, Japan, and Taiwan.
The U.S. National Geospatial Intelligence Agency manages the Maven program, which is an AI-based computer vision software that is used to search video and other image sets for targets for warfighters. Jaskolski said Maven was a “great idea” when it began but involves a tremendous amount of data labeling.
“But as it goes on, people are, I believe, starting to realize that you can’t label your way out of an AI arms race with China,” he said. If it comes down to out labeling China, I think we might lose that battle because there are more people, more energy that can go into that.”
Eventually, Synthetaic wants to get its technology to the edge, which will reduce latency and speed the time from query to results. For example, Jaskolski said, if data could be ingested directly into RAIC from a satellite base station an analyst could ask whether there are new concrete pads in North Korea and results would be ready within minutes.
“It’s almost like a deep search across all of the national technical and commercial data and effectively, nothing could hide,” he said. “You’d have complete, effective visibility into all of the EO (electro-optical) and SAR (synthetic aperture radar)-based geospatial activities on Earth.”
RAIC is part of an edge application offered by Booz Allen called Big Horn, which is a ruggedized kit used to refine AI models in minutes in disconnected environments.