There were hoodies. Robots. Free drinks. Young founders milling around a loftlike, concrete-floored space. Your typical startup demo day, in other words--except for the presence of a four-star Army general and pockets of uniformed military personnel and besuited corporate types with name tags from giant defense contractors like Booz Allen Hamilton. And the senior U.S. senator from Texas, John Cornyn, standing near the general, along with the mayor of Austin and the speaker of the Texas House of Representatives.
It was February 21, opening day of the Center for Defense Innovation, in the downtown Austin high-rise that houses Capital Factory, the city's leading accelerator. Last fall, the U.S. Army chose Austin as the home for its new Army Futures Command, the most sweeping modernization effort in decades across the largest branch of the military. As the pace of tech change has quickened, the Army has begun to look badly outdated in fields like artificial intelligence and robotics. The AFC intends to fix that by working with startups; the new space within Capital Factory is where it all will come together.
"This is the only place in the world where an entrepreneur can walk in off the street and engage authentically with members of the defense and intelligence communities," said Capital Factory founder Joshua Baer in his remarks. "The best petri dish for innovation in the nation," Cornyn intoned. Besides the AFC, several other military-innovation groups--notably the Air Force's Afwerx and the Pentagon's Defense Innovation Unit--were establishing beachheads in the new space, where they'd spend time with scrappy founders, sharing ideas, spreading around some of the $320 billion the Pentagon spends annually on contractors, and generally helping to keep the peace, or at least ensure U.S. dominance.
It sounded good. It looked good--disparate worlds coming together with a shared purpose. It was happening, though, just as there was much handwringing in tech about working with the military. Last June, Google canceled its Project Maven partnership with the Pentagon--it sought to improve drone-strike accuracy with A.I.--after about 4,000 employees protested. The very week that the Austin Center for Defense Innovation opened, more than 100 Microsoft workers signed a letter to their top executive objecting to a $480 million Army program to turn the company's HoloLens augmented-reality headsets into battlefield devices intended to increase soldiers' "lethality." (Microsoft CEO Satya Nadella stood firm, citing a patriotic duty to work with the military.) But beyond the ethical issues of turning commercial products into killing machines, for startups, there's the brute fact that the government is often a lousy partner, with its slow-grinding pace, its layers of bureaucracy, and the potential fallout from changing political winds.
Still, hundreds of Austin entrepreneurs have shown up for defense-focused events at Capital Factory. Among the companies that demoed technology on opening night: Valkyrie, whose machine vision may help detect potential explosive devices; Apptronik, which showed off a pair of bipedal robot legs; and Athena Security, which spots firearms in a crowd. And then there was Senseye, founded by David Zakariaie, which makes technology that enables computers to read people's minds by monitoring their eyes.
Zakariaie, who has dense stubble and close-cropped black hair, is just 22. A Disney freak who visits its theme parks every month or two and can recite endless arcana about them, he speaks with a boyish stutter that makes his eyes flutter and tends to make listeners hang on his every word. That night, he huddled with government officials while staffers administered a lie-detector test to attendees: A camera peered into subjects' eyes as they answered questions, and a digital readout indicated their truthfulness as measured by tiny ocular movements--allegedly with far more accuracy than a polygraph. Zakariaie had conducted countless such demonstrations in recent months, as local leaders paraded him in front of military brass and media every time there was an event related to the AFC. "The poster boy," Baer called him.
Senseye had already landed a contract with the Air Force. But what few besides Zakariaie knew that night was how difficult that relationship had become. Being a mascot for startups working with the military is one thing. Making that odd-couple collaboration work, he was learning, was a much different story.
At times, it seems some mythmaker conjured up Zakariaie as the prototypical wunderkind teen with a world-changing idea and no internal limiter. He was an awkward 15-year-old when he hatched the idea that became Senseye. Back then, he lived in Los Angeles and spent his afterschool hours tinkering in a UC Irvine nanotechnology lab. He decided to try to build a microscopic computer--using DNA instead of silicon--small enough to fit inside, say, a contact lens. ("This technology carries the implications and potential to be a total game-changer in the sense that DNA computers will render current silicon-based technologies as useful as the cassette tape," he wrote in a 2016 patent application.) He put together a proof-of-concept paper and submitted it to the California State Science Fair--and was promptly disqualified because, he says, the judges didn't believe he'd done the work himself. But an admiral from the Office of Naval Research approached him after the awards ceremony and offered him a small grant to pursue his vision.
Zakariaie tested out of high school, got to work, and soon his idea morphed. "As awesome as the bionic chip concept was, I realized it didn't make sense until we figured out the command-and-control mechanism around it," he says. "If you had a computer in a contact lens, you're not going to use a keyboard or a mouse or a touchscreen. Those things serve as a bridge between man and machine. So the idea was to shift it--to figure out a way to tap into the brain wirelessly."
By late 2013, when the world was agog over Google's connected eyeglasses (remember when Google Glass was a thing?), Zakariaie began homing in on how his idea might become a business. He would build market-research applications for Glass, he decided, that collected data on everything shoppers interacted with in a store. But he realized "the data would be worthless until we could figure out why people made purchase decisions. Did you buy the Tide detergent because of the price? The packaging? The shelf placement? Because your mom always bought it?" Decades of research had shown that reading pupil dilation provides important clues to brain activity, but this works in the lab better than in the real world, where it's difficult to isolate what a person is reacting to. Zakariaie decided to focus on the iris and its thousands of muscle fibers that control the pupil and tie back to the nervous system. Able to measure much finer eye movements, he could, in theory, begin to understand more precisely what drives them.
In 2015, he raised six figures worth of funding from friends and family, hired a couple of neuroscience PhDs from UCLA--and promptly burned through all the cash and laid off his six-person team the following year. "You give a bunch of money to an 18-year-old who has no idea what he's doing and that's what happens," Zakariaie deadpans.
Desperate to find a lifeline, he heard about a startup competition in Austin in early 2017 that would award its winner $100,000 and office space at Capital Factory. A few weeks later, Zakariaie landed in Austin at 11 in the morning, pitched at 12:30, won the competition at 4, and was flying back home before dinner. Two weeks later, he moved from his parents' house to Texas.
His timing was good. Baer has long been a key cheerleader for Austin's tech scene, and having a precocious and wildly ambitious young founder at hand was perfect for his narrative that Austin was becoming the next great tech hub. President Obama's secretary of defense, Ash Carter, had made a couple of visits to Capital Factory the previous year and left intrigued, and then various organizations within the DOD began visiting Austin, seeking ways to work with startups. Baer started bringing Zakariaie around to meet the military brass, thinking they might find mind-reading technology compelling.
Zakariaie was grinding through 18-hour days at Capital Factory--tearing through neuroscience research, looking for investors, looking for customers--so he was more than happy to perform as Young Einstein for visiting VIPs. The exposure? The potential for doing business with the U.S. government--one of the biggest customers on the planet? Yes, please. "I'll be your poster boy all day long," he told Baer.
The U.S. military has a long history of trying to engage innovators from the private sector and academia (think Darpa, for instance). But those efforts are notorious for work that never crosses the gap--known in defense circles as "the valley of death"--from R&D to actual production. Most big Pentagon contracts have long gone, instead, to a few giants--Lockheed Martin, Booz Allen Hamilton, Raytheon--that specialize in defense work. As Army Lieutenant General Paul Ostrowski, who directs the U.S. Army Acquisition Corps, has put it, only 5,000 U.S. companies do business with the DOD--out of the six million U.S. companies that have more than one employee. But that's mostly been fine with everyone. The big defense contractors--the "Beltway bandits"--know how to navigate bureaucracy. The DOD knows what it's getting. The pace, while plodding, has been effective.
But even the bureaucracy knows that this way no longer works. The Army's approach to tech has largely been "looking into the future 10, 20, 30 years, and making bets on what we thought it would look like," says Adam Jay Harrison, the AFC's newly appointed command innovation officer. "That ain't the world we live in today."
Harrison, a former Naval intelligence officer, has a vibe that skews more biker than military: Stout and bearded, he wore a black Adidas tracksuit and a chunky metal bracelet the day I met him in his black-walled office. He's setting up a concierge service of sorts within the Futures Command, to help startups navigate the process of doing business with the Army. Harrison left the public sector in 2006 to start Mav6, a big-data defense contractor that made the Inc. 500 two consecutive years before being sold. After learning the contracting process from both sides, he set out to fix it, conducting a study at NYU that proposed a modernization plan. It's now a rough blueprint for his new job.
Technology moves too quickly today, Harrison says, for the DOD to expect its core contractors to keep up. And many of the millions of companies that don't work with the military are making things that the DOD could use. Meanwhile, Chinese and Russian tech is catching up with the U.S.'s--and those countries have some built-in advantages. "China can reprogram its commercial industry whenever it wants to support national objectives," Harrison explains. But the U.S.'s bottom-up approach means a bigger bounty of disruptive new ideas. "For the DOD to remain competitive, we have to harvest that activity," he says. "Right now, we make it really difficult for that to happen."
But contracting with the government can mean miles of red tape--"180,000 pages of stuff," in Harrison's words. "It gets pretty difficult for a small business to manage," he says. There are workarounds that the Defense Department can use to get money more quickly to "nontraditional contractors," he explains, and he's employing them as much as possible. More problematic, he says, is that "startups don't really understand what we're looking for."
Fixing the latter means more than sharing wish lists-- it requires making government work appealing, not a slow grind. Harrison plans to focus on what he calls dual-use opportunities, where the DOD acts as a starter market for products that companies can also sell to others. That's not altruistic; it can make prices come down with scale, so the military can get what it needs affordably. Each of the military innovation units hopes to create similar win-win scenarios. But even Harrison refers to many DOD efforts to work with startups as "innovation theater." And Trae Stephens, partner at the powerful Silicon Valley venture capital firm Founders Fund and chairman of Anduril Industries, a tech-focused defense contractor that he co-founded with Oculus Rift creator Palmer Luckey, says the Pentagon's concept of making small investments in many projects is inherently flawed: The process often starts with a one-time payment to test an idea, followed by rounds of one-offs as that idea gains traction, before a startup can land a contract that ensures revenue. The chances of making it to that final stage are small--and the Beltway bandits often elbow their way in, grabbing contracts and then subcontracting to smaller companies. For most startups, Stephens says, it's a lot of work for little money and a poor shot at success.
"To be successful, you have to build your whole business around defense," Stephens says. "Or at least wait until your business is way more mature and you can have hundreds of people working on it." Neither, of course, describes Zakariaie and Senseye.
In June 2017--just weeks after Zakariaie arrived in Austin--Capital Factory's Baer suggested that the young founder give his lie-detector demonstration to two Air Force officials. Zakariaie jumped at the chance. He got his pilot's license at 19 and is now a flight instructor, the highest level of pilot certification. ("I just decided to go all the way," he says.) During that meeting, as Zakariaie explained how Senseye measured "cognitive load"--how hard a subject's brain worked as he or she answered a question--an Air Force official cut in: Could Senseye measure cognitive load while someone operated a flight simulator? The Air Force was looking to speed up its flight training process, he explained. It might be more efficient if a computer could sense in real time whether the student had achieved mastery of a subject, and, if so, cut the drills short.
Senseye could, Zakariaie said. The company wasn't even planning on applying eye-monitoring tech to education, but Zakariaie saw the potential "almost immediately." More efficient training would be a big deal in so many fields. And having a branch of the U.S. armed forces using his product would be a powerful persuader to other customers. Even if the Air Force ended up as the only client, he calculated that flight training could be a $20 million or $30 million business.
By that November, Zakariaie was testing that product, and in February 2018, the Air Force began its Senseye-powered Pilot Training Next (PTN) program with a 30-person class of pilot trainees donning virtual-reality headsets that monitored their eyeballs as they flew simulators. Training that normally lasted more than a year took six months.
But the experience was a slog. Rather than work directly with Senseye, the Air Force hired the company as a subcontractor to a multibillion-dollar Virginia-based defense contractor, SAIC. Zakariaie then had a team of three, no one to help with administrative tasks--and roughly 40 lengthy forms that he had to fill out, including multipage lists of laws to research and agree to comply with. "Our financials had to be done a certain way. There was all this hyper-specific information they needed," he remembers, his exasperation still fresh as he scrolls again through those lists of laws on his laptop. For 10 days, he worked on nothing else.
It seemed worth it for a good-size check--more than a million dollars--and the potential for far more if the program succeeded. So, last fall, Senseye agreed to participate in a second-round PTN trial, on the condition that it could contract directly with the Air Force. But, as the program began in January, Senseye still didn't have a contract. No word came until mid-March--almost halfway through the program--when, Zakariaie says, he was about to take the stage to lead a panel at Austin's SXSW festival and an Air Force official pulled him aside and said, without explaining why, that the new contract would be for $250,000--a nearly 80 percent reduction in the fee.
Zakariaie was aghast. He'd shelved other priorities to focus most of the team's energy on flight training--and he'd hired aggressively: The company now had two dozen employees. He'd counted on the renewed Air Force deal to keep making payroll--and never imagined the government would essentially stiff him. But he gamely stepped onstage and went ahead with his panel--one called "The ABCs of Securing a Government Contract."
He'd scheduled a birthday trip to Disney World right after SXSW. There, between rides on the Frozen roller coaster, he drafted a letter to the Air Force informing it that Senseye was pulling out of the pilot-training program, a move that would effectively end any relationship with the service. (He "let it go," he cracks.) Upon returning to Austin and talking it over with his COO, Sam Marshall, an Austin tech veteran he'd recently hired, Zakariaie sent the letter, after which Senseye remotely disconnected its software from the Air Force's flight simulators. (An Air Force rep would not discuss any details beyond confirming that the Air Force worked with Senseye on two rounds of testing, and that the second ended prematurely.)
Then Zakariaie and Marshall scrambled, reassuring their team, hustling to lock in new business. Zakariaie had hoped that, after a second successful round of flight training, he could raise a funding round without giving up too much equity. Now he had to raise sooner, from a weaker negotiating position--and his and Marshall's recent conversations with venture capitalists had been mostly disappointing.
"The word government carries so much baggage with investors that it kind of obscures the exciting stuff," Marshall explains. "They're like, 'What else do you have? The government is known to cut budgets unexpectedly.' "
Zakariaie left for London to "sell the shit out of" a program Senseye was developing with Lloyd's Register, a multinational risk-management company that was started more than 250 years ago. (It's unaffiliated with the insurance giant Lloyd's of London.) For Lloyd's, Senseye was planning to develop an eye scanner to test whether container ship crew members are safe for duty, and not tired or drunk or distracted. Back home, Marshall reflected on the recent whirlwind, with a better understanding of investors' concerns. "As a startup, you have to be very diligent about where you focus your resources," he says. It's not that pilot training was too far afield--he still finds the potential for educational applications of Senseye's technology alluring. But, he concludes, "just because something works doesn't mean it fits."
As the Air Force deal collapsed, Zakariaie entered discussions with Harrison and the Army Futures Command--to re-create the Air Force training program for Army aviators. That partnership hadn't been signed as of presstime, but the hope is for a six-month engagement. Harrison sees that as a test case for a bigger idea: "You could put that sensor on the top of every classified network," he says. "So when somebody is logging into their computer in the morning, they have to answer a question: Have you stolen or misappropriated classified information? It would fundamentally change how we deal with classified materials."
He sees Senseye as an "archetype" for the businesses Futures Command wants to work with. "We want companies that didn't wake up in the morning thinking, 'How am I going to help national security?'--companies that have an idea we can relate back to our needs and then be an enabler for their business."
Senseye has gamed out a multifaceted best-case scenario with the Army that, when they're feeling optimistic, Zakariaie and Marshall think could eventually be worth as much as $100 million. But there's a snag: The AFC was hoping to use data from Senseye's Air Force program to validate its tech; since that deal was cut short, the Army must fund its own validation before making any commitments.
But a more familiar deal might get Zakariaie closer to his dream of controlling microscopic computers with the mind. During SXSW, he impressed investors from Japan's SoftBank, which had just started a new venture capital fund--Deepcore--that focuses on early-stage advanced-technology companies. Once word spread that SoftBank was leading a seven-figure investment round in Senseye, other VC firms got excited--including several that had turned Zakariaie and Marshall away. And Zakariaie says he'd work with the military again--even the Air Force. "We're treating this as a lesson learned," he says. "All the processes they are trying to deal with, we experienced."
Should Startups Work With the Military?
Yes. It's your duty. When Microsoft CEO Satya Nadella faced an employee protest about a nearly half-billion-dollar contract with the Army, he said it had been the result of a "principled decision...to protect the freedoms we enjoy." It's a patriotic perspective shared by many tech leaders--even when it comes to building tech that enhances weaponry. "At the end of the day, we use drones. We use drones to kill people," says Senseye's David Zakariaie. "If the drone needs to identify if someone is a threat or not before we pull the trigger, we better have the best A.I. possible."
No. Not unless you're all in. Trae Stephens, a partner at Founders Fund and co-founder of defense contractor Anduril Industries, sees most startups as a bad fit with the Pentagon. "It doesn't work for a mostly commercial business," he says. "Your people are not ideologically aligned with that kind of work. That's when you get Google pulling out of Project Maven," the controversial program to improve drone-strike accuracy. So, what's the solution? "We need to build new prime defense contractors," he says, "so people can work knowingly on military applications. We won't get the best defense-relevant tech out of commercial industry."
Maybe there's a third way. New York City-based A.I. startup Clarifai experienced both the peril and the promise of working with the Pentagon when it agreed to help out on Project Maven. While some Clarifai workers protested, the company also recognized a promising line of business. Last December, Clarifai announced a new division called Neural Net One, based in Washington, D.C., that would focus solely on defense work while the rest of the company remained dedicated to commerce.