Deepnight co-founders Lucas Young and Thomas Li have been friends since childhood. Both were working as software engineers at Google when Young decided he wanted to crack the code, so to speak, on a problem that had plagued the U.S. military for decades: digital night vision tech.

Most night vision technology is still analog. Goggles use optical lenses and a chemical process to convert the scant light at night into images, Young told TechCrunch. And they cost from $13,000 to $30,000 apiece from military contractors like L3Harris and Elbit America.

For years, the U.S. Army has been trying to digitize the tech, mostly focused on the hardware. A case in point: the $22 billion budget for the Integrated Visual Augmentation System (IVAS) project that Anduril just took over from Microsoft and its Hololens tech.

Young, who has a degree in computational photography from CalPoly, spent five years working on smartphone camera software. He wrote code that offset the limitations of the tiny aperture, cheap $50 digital cameras used in smartphones. And Li’s background is in AI tech, particularly computer vision.

One day, Young read a scientific paper from 2018 called Learning to See in the Dark co-authored by well-known scientist Vladlen Koltun, who is currently at Apple. It discussed using AI for low-light imaging, but at the time, on-device AI chips weren’t fast enough to support the 90 frames per second (fps) necessary for realtime viewing.

In 2024, Young realized that AI accelerators running on System on Chips (SoCs) had advanced enough to support 90 fps. He talked his friend Li into quitting their jobs and founding a startup called Deepnight. And they promptly got into the Y Combinator winter cohort.

The military was their first obvious customer, but they couldn’t just roll up to the Pentagon and book a meeting. So, Young found an industry event where people from the U.S. Army’s night vision laboratory were in attendance.

He wrote a white paper that outlined his idea: night vision as a software problem. He handed out copies at the event, including to an army colonel who agreed to read the paper. “It was just a hallway conversation. I wasn’t even in business attire. Just a T-shirt,” Young recalls.

The colonel liked what he read enough to put the founders in touch with people at the lab, formally known as the US Army C5ISR Center.

Desperate to show those folks that their concept would work, the founders built a night vision smartphone app. They put the smartphone into a smartphone-holding VR set.

It was a rudimentary prototype that was impressive enough enough to lead to their first sale.

“The army awarded us a $100,000 contract in February 2024, one month into Y Combinator, based on the proof of concept in a smartphone demo and our whitepapers and presentation,” Young said.

Young and Li then had to present their progress in a more formal demo. The pair flew to Washington D.C. to show a room packed with 10 people how their software worked as well as state-of-the-art goggles, Young said. (Here’s a YouTube video where they demo their tech.)

The meeting led to more contracts. A year after launching, the startup has booked about $4.6 million in contracts from the federal government, including the US Army and Air Force, as well as with companies like Sionyx and SRI International.

Deepnight also promptly attracted investors. By the end of YC, it raised a $5.5 million round led by Initialized Capital, with angels like Kulveer Taggar, former In-Q-Tel partner Brian Shin, and Matthew Bellamy, lead singer of the band Muse. Y Combinator also chipped in their standard deal.

Perhaps best of all, Koltun, the scientist who wrote the paper that inspired the company, also became an angel investor.

Deepnight offers software and partners with hardware makers like goggle manufacturers or military helmets or other products.

“Now we can make everything in the world see in the dark, because it’s just a software program. So that’s automotive, security, drones, maritime like boats, electronics, nav cameras,” Young describes. And because it all relies on an off-the-shelf $50 smartphone camera, their tech doesn’t need expensive bespoke hardware.