A new hope (for pilots): Beacon AI Series A raises $15m for pilot assistance technology
Picking up distress signals, engaging and disengaging autopilot, fixing bugs in-flight; these are all tasks R2-D2 completed in the Star Wars movies. Without the likeable droid’s help, the story would have been a lot different (definitely shorter).
As a child, the idea of a real-life R2-D2 or C-3PO was the stuff of fantasy imagined by my cousin and I as we rewatched A New Hope for the umpteenth time. But now, with 10 signed Department of Defense (DoD) contracts, an 11th selection and several airline engagements, including with Emirates Airlines, one Californian startup is turning those ideas into reality for pilots.
Founded by former US Navy pilot Matt Cox, Beacon AI is on a mission to put an artificial intelligence (AI) assistant or copilot, aka R2-D2, in every professional flight deck. The startup, founded just over three years ago, today announced the completion of its Series A funding round, raising $15m. Oversubscribed in terms of interest, the round was led by Costanoa Ventures with participation from Scout Ventures, joining existing industry heavyweights Sam Altman and JetBlue Ventures. The Series A brings Beacon’s total investment to $20m.
“There is no better team to bring AI to the flight deck for commercial and military pilots,” said Greg Sands, founding and managing partner at Costanoa. “They have already proven themselves in the most high-stakes cockpits in the world, which is why we’re excited to see their progress with commercial pilots.”
Founder and CEO Cox (pictured below right) has deeply personal reasons for pursuing the development of Beacon AI’s technology. As a self-described “average at best” F-18 fighter pilot flying on and off aircraft carriers, he had his own challenges, as well as losing numerous friends, including the instructor pilot he first flew with in an F-18.
“Flatly, it is a product I wish I had when I was flying in the Navy,” Cox tells us. “I’ve known quite a few people over the years who have died, and when I look at a lot of the near misses that are in the news today, at least the non-mechanical ones, the vast majority of them are potentially preventable human error.”
“As humans, we can all relate to entering the time on the microwave wrong or entering the PIN into our phones wrong. While in regular life, those are pretty inconsequential items, in a cockpit, that level of mistake can lead to very real consequences.”
Cox references the crash involving Asiana Airlines Flight 214 on final approach into San Francisco Airport on July 6, 2013. The Boeing 777-200ER in question stalled and crashed with 307 people on board, three died and another 187 were injured, some seriously. He says the incident was due to human error and a “misunderstanding of the system.”
After leaving the Navy, Cox worked at, among others, a self-driving car company Cruise and the DoD’s Defense Innovation Unit. Upon leaving Cruise in 2020, the veteran pilot was not convinced by the zero-pilot-first approaches being taken by most of the players in airborne autonomy.
“I felt the answer was to take the best parts of what advanced software and computers are good at and pair it up with the best parts of what humans are good at,” he says. “That is our why; we feel there is an obvious gap in the system.
“Airplanes have been remotely piloted as far back as 1917 by Archibald Low, and further in the 1930s with the Jenny. Whereas we think the primary gains will be through better data, leading to improved software and decision-making. For us, that means taking the best part of the machine and pairing it with the best part of the human. We are aimed at the professional pilot with the intention of augmenting professional aircrews.”
So, how does it work? First and foremost is the data, which Cox describes as a content- and context-rich dataset similar to that used by the pilot. The problem is that the data set doesn’t exist, which means the Beacon AI team is building the platform to create it. With that data in place, Beacon can then incrementally deploy features to assist aircrews during simulator and aircraft flight operations.
“You can think of it like advanced driver assistance systems [ADAS] for cars, which include features like automatic emergency braking or adaptive lane following, and over time, it keeps getting incrementally better,” says Cox. “Of course, I’m aware that cars have some unique challenges, like a really complicated perception environment and short reaction times, which makes having a human in the loop very difficult. On the other hand, airplanes typically have a longer reaction time and a less complex perception challenge. I think in many ways the ADAS style approach will be better for airplanes than in cars.”
The startup has just doubled the team size to meet growing interest from civilian and defence customers. The team is expecting to grow significantly and is hiring predominantly in the SF Bay Area, says Cox. Today Beacon AI is made up of a combination of former military and commercial pilots and autonomous driving engineers. Members include ex-Navy fighter pilot Cameron Douglas, who handles business and product development, and senior software engineer Rohan Deshpande, who previously spent four years at General Motors, including several years at autonomous driving programme Ultra Cruise.
Beacon co-founder Avinash Nair told us: “At Beacon, we’re creating technology that has the potential to transform flight safety, and it’s inspiring to know that our innovations could help prevent incidents we often see in the headlines. Being part of a team that directly enhances passenger safety and shapes the future of aviation is truly rewarding.”
Earlier this year, Beacon announced the completion of a Phase 2 prototype Other Transaction Agreement (OTA) from the US Special Operations Command. The OTA focused on advancing human-machine interactions intended to address aspects like attention, fatigue and physiological risks. Pilots’ physiological risks increase on substantially longer DoD missions and commercial flights, which this OTA and follow-on work aims to assist with.
Cox explained: “I had the opportunity to work with and fly with the DOD and recognise that they have additional challenges in their operational environment. Many of those challenges are also present in commercial aviation, where fatigue remains a major risk factor. The capabilities we are adding and enhancing will help pilots address hazards, including standard human error and cognition, airborne hazards and fatigue, while advancing the state of the art for human-machine teaming.”
The DoD sees aircrew augmentation technology like Beacon’s as a pathway to enhance operations and create battlefield advantages, especially in the face of much longer and potentially complex Pacific region missions. “Solutions that are equally pertinent to the commercial market,” says Cox.
Beacon AI’s founder likes to think of his company as a “bridge to the inevitable future.” The last 120 years of aviation technology puts most of the burden of complex data handling and decision-making almost exclusively on the aircrew. Any future solution must be able to join the crew and assist by presenting the most relevant information, reducing the possibility of human error and simplifying complex tasks. Cox believes in the near future, humans won’t have to do all of it on their own.
“Beacon’s job is to pull the inevitable future to the present, or as close as we possibly can. We think there are many procedural, routine roles within the flight deck that we will do very well, and there are other parts which are very complex and do not have good training data that we think are best deferred exclusively to the crew,” explains Cox.
Beacon AI is approaching its first external deployment. But when asked about the timeline for when the system will be fully ready, Cox has a different approach. He emphasises that while it’s challenging to predict exact outcomes, Beacon AI’s focus is on building practical features that address current challenges. As it gathers more data from external users, the system’s learning and capabilities will expand exponentially.
That means Cox is optimistic about the future, seeing Beacon AI as a tool to enhance pilot decision-making in the near term, while acknowledging the importance of a thoughtful, iterative approach to developing the next generation of human-machine collaboration.
Subscribe to our free newsletter
For more deep dives from SAF Investor, subscribe to our email newsletter.