AeroVect taught the airport to drive itself.

Scene One

The tractor with no one in it.

It is 4 a.m. at a major U.S. airport. A baggage tractor pulls a chain of carts away from a parked widebody, banks neatly around a fuel truck, and threads itself back to the sort facility. There is no driver. No remote pilot crouched in a trailer. Just a hardware-agnostic stack of lidar, cameras and code, mapping the tarmac in real time and behaving, somehow, like a competent grown-up. That stack is called the AeroVect Driver. The company that built it is a five-year-old Silicon Valley startup with fifty-eight people, two Harvard co-founders, and a contract to put autonomous tugs to work in Dubai, Atlanta and most of the places in between.

Scene Two

The problem no one books a ticket to see.

Aviation has a labor problem and a margin problem, and both live below the wing. The work of moving bags, cargo, dollies and tugs across a tarmac is unglamorous, weather-exposed, hard to staff and, increasingly, hard to keep safe at scale. Ground handlers - the contractors who do this work for airlines - run on thin margins and high churn. When staffing wobbles, flights wobble. When flights wobble, the airline gets blamed. When the airline gets blamed, the airport hears about it. Nobody in this chain has a spare hour.

The obvious fix is autonomy. The unobvious problem is that, until recently, nobody seriously tried to solve it. The autonomy money went to passenger cars, where humans are unpredictable and lawyers are everywhere. The ramp, by comparison, is a controlled environment with a fixed map, repeated routes, professional operators, and an organization that can actually procure new equipment. It is, in other words, a self-driving founder's dream. It just doesn't make for a good TED Talk.

Caption. The world's busiest ramps look like a slow ballet performed by very serious men in reflective vests. AeroVect would like to give half the dancers a coffee break.

Scene Three

The founders' bet.

Raymond Wang and Eugenio Donati met at Harvard. In 2020 - a year, you may recall, with limited social calendar - they started building autonomous airport vehicles out of a garage. The early Xfund check was, by the firm's own admission, under $200,000. The bet was simple and slightly stubborn: most autonomy companies were going wide. AeroVect would go narrow. One environment, one customer profile, one problem worth charging real money to solve.

The team they assembled did not look like a typical airside vendor. Engineers from Argo, Apple, Ford, Stanford and MIT signed on. The product was built for the realities of the ramp - dust, jet wash, GPS multipath off metal aircraft - rather than for a freeway in Mountain View. Investors followed. Graphene Ventures led a round. In April 2025, Future Back Ventures joined a seed extension. Total capital raised to date sits in the high tens of millions, which, by autonomy standards, is alarmingly disciplined.

Scene Four

The product, demystified.

The AeroVect Driver is a retrofit autonomy stack. A 360-degree perception system - lidar, cameras, high-precision GNSS from partners like Point One Navigation - sits on top of an existing OEM tractor. The software fuses the sensors against a high-definition map of the airport, classifies what it sees (aircraft, vehicles, humans, dollies, jet bridges), and decides what to do next. Crucially, the stack is hardware-agnostic. Charlatan's autonomy is brittle to one specific chassis. AeroVect's is not. The same brain runs on different OEMs because, frankly, the airline doesn't want to rebuild its fleet to use it.

Translation, for non-engineers. It is a self-driving kit you can bolt onto the tractor you already own. The tractor does not need to know it is famous now.

Scene Five

The proof, in numbers.

Pilots are easy. Production is the bar. AeroVect crosses it: their tractors are not a slideshow at a conference - they are doing live work, around live aircraft, with live consequences. The chart below is the ramp story compressed into bars.

Scene Six

The mission.

AeroVect's stated aim is to make airside operations safer, more reliable and more cost-efficient. Strip the press-release language, and the substance is this: the ramp should stop being the fragile, labor-bound part of aviation. Bags should move on time. Tractors should not crash into jet bridges. People should be redeployed to work that requires judgment - not to driving the same eight-minute loop in the rain at 3 a.m. The company is hardware-agnostic on principle because it wants the upgrade to reach every airport, not just the new ones.

Caption. The romantic version of aviation is the cockpit. The economic version is the tug. AeroVect is fixing the economic version.

Scene Seven

Why this matters tomorrow.

Air traffic keeps climbing. Ground handling staffing keeps not climbing as fast. Insurance keeps getting more expensive. The bet that wins this decade in physical AI is not the bet that automates the most glamorous environment - it is the bet that automates the most leveraged one. Airports are leveraged. A 15% efficiency gain on the ramp ripples into on-time performance, into airline cost per available seat mile, into passenger experience, into the share price of carriers you can name without thinking. AeroVect is not in the business of changing how flying feels. It is in the business of changing what it costs to make flying happen.

Closing Scene

Back to 4 a.m.

Return to the tractor. It is still moving. The widebody it just serviced is being pushed back. Somewhere, an operations dashboard logs the crossing as one more among tens of thousands. The driver's seat is empty. The shift supervisor, who used to spend her night chasing tug coverage, is reviewing a fleet view from inside. The airline customer hasn't called. The plane left on time. The tractor turns around and goes to do it again. It is, by any honest definition, a small revolution. It is just not the one anyone was watching for.