Up-and-coming software company Picogrid has won a $9.3 million Small Business Innovation Research (SBIR) Phase III contract, the small firm’s largest to date, to connect disparate counter-drone defenses at selected Air Force bases using “translator” software written by AI.
The translator is called Legion, and it connects hitherto incompatible military systems — for example in this contract, the different radars, jammers, and other anti-aircraft systems at an air base. That requires a careful, multi-step process for each system added, starting with a thorough analysis of its technical specifications and ending with a custom-coded translation module. Human engineers could easily take weeks to integrate a single weapon system.
The announcement declined to specify at which Air Force bases Picogrid’s software will be integrated.
“When we first started [integrating] our first one or two platforms, we’re looking at about a month, month and a half, to actually carry out these integrations,” Picogrid’s cofounder Martin Slosarik told Breaking Defense ahead of the company’s announcement. With the new AI, he went on, “we’re looking at less than a day for most of these platforms.”
Sometimes it’s much less: “Last week, we had a company hackathon where we actually implemented seven different integrations in about 70 minutes,” he said.
“It’s something that really wouldn’t be possible 12 months ago,” Slosarik said. This is progress from last June, when he compared the AI coders available at the time to an entry-level human software engineer that needed heavy supervision. Now AI’s increasingly able to do the whole job itself, he said.
“When we spoke during the summer, I told you, ‘hey, we’re able to use AI to deliver these integrations with these different companies building sensors, robotics, drones, what have you, and AI takes it 60 percent of the way, and our engineers take it the rest of the way,’” Slosarik said. “Well, now we’re really at, I would say, 95 percent of the way, 98 percent.”
Obviously, it takes longer than a day to deploy this code to an operational military system and test it to the armed services’ satisfaction. Leading up to this latest contract, Slosarik said, “we’ve been sort of fielding and building … experience for the last year and a half with the Air Force.” The project got new impetus after Ukraine’s Operation Spiderweb drone strikes last June took out a third of Russia’s strategic bomber fleet at its own bases, he noted.
RELATED: Pentagon expands counter-drone authorities beyond ‘fence line’
Picogrid says it’s been able to build on those experiences to turn around other integration projects much more rapidly. Late last month, the company announced it had integrated five different sensors from five different companies into a coherent drone-detection system for a 1st Cavalry Division unit at the National Training Center on Fort Irwin, Calif. The disparate active radar, passive radar, and acoustic sensors were then able to feed data into the Army’s standard command system for battlefield air defense, Northrop Grumman’s FAAD-C2 (which itself is getting AI features). The software’s even smart enough to use the passive sensors to get an initial fix and only turn on the active radar — which emits a highly targetable signal — briefly for confirmation.
Picogrid did that project from a cold start in about four weeks, beginning with an unexpected email from 1st Cavalry Division leaders wrestling with a bevy of unfamiliar counter-drone technologies that didn’t work together.
“1st Cav came to us about a four-week notice before they went into the ‘box,’” Slosarik said, using Army slang for the National Training Center’s thousand square miles of rugged desert — a notoriously difficult environment for humans and technology alike.
Slosarik sees the $9M Air Force contract and the 1st Cavalry quick-turn as models for an ambitious challenge to established defense integrators, touting equal collaboration among companies through the Picogrid Partner Ecosystem, currently at 50 members, as an alternative to traditional uneven partnerships between big primes and smaller subcontractors.
“We’ve been really picking up our speed: We’re at just short of 100 different integrations now, and it’s really just possible because of that rapid increase in the speed of AI-driven development,” Slosarik said. “Traditionally, you have these large players selling butts in seats to integrate different programs and systems. Well, frankly, with the right type of infrastructure and the right type of engineers, we are moving towards a world where the speed and cost of the integration asymptotically approaches, you know, zero.”