Login
Earth first, Mars later: Inside AIM’s grand vision for physical AI and autonomous bulldozers
In a headquarters and lab space formerly occupied by SpaceX in Redmond, Wash., AIM Intelligent Machines (AIM) is focused on solving big problems on Earth. But the startup’s CEO envisions a day when autonomous bulldozers and excavators will dig, haul, and grade on the moon or Mars, and take AIM’s “terraforming mission” off planet.
For now, AIM’s 25,000-square-foot facility in a nondescript business park is a long way from Mars. Inside the sprawling space, there are glimpses of what the rapidly growing company is working on, including the apparatuses that attach to existing machines to make them self-driving.
Around the office, desk cubicles are decorated with tiny yellow excavator buckets, mirroring photos on the walls of heavy equipment operating on job sites worldwide.
The toy excavators are a nod to a massive global market that AIM founder and CEO Adam Sadilek wants to continue to disrupt with modern technology.
Autonomous passenger vehicles have captured the public’s attention for decades, but construction, mining and hauling equipment attracts little fanfare, even as legacy companies including Komatsu and Caterpillar embrace new technology.
AIM’s goal is not to build new machinery, but retrofit existing earthmoving fleets with a physical AI platform — using advanced sensors and edge compute to let heavy iron operate entirely on its own.
Founded in 2021, the startup grew out of Sadilek’s background at Google where he spent nine years working on projects involving AI and autonomous vehicle systems.
Whether building anti-flood structures or wildfire breaks, managing nuclear waste, mining critical materials or clearing land for agriculture or the military, Sadilek views AIM’s work as immediate terraforming on Earth that is necessary to drive down costs for housing and commodities. But the long-term vision remains interplanetary.
“When humanity goes to Mars, the real question is not so much around what the rocket looks like as a vehicle to get us there, but what is going to happen after the rocket lands,” Sadilek said. “You cannot have human operators run there. That’s why this is a very long mission that we are on.”
Building autonomy for heavy equipment presents a paradox self-driving cars never have to face: the ground itself is constantly changing. While a Tesla or Waymo relies on pre-mapped roads and predictable lanes, a bulldozer or excavator’s entire job is to reshape its environment. AIM’s physical AI platform has to continuously build real-time 3D maps using onboard 360-degree LiDAR and edge compute, making split-second decisions without relying on persistent GPS or cloud connectivity on remote job sites.
Furthermore, taking human operators out of cab seats addresses one of the most perilous aspects of heavy industry. By creating “zero-entry” sites where machines operate autonomously, AIM’s platform effectively removes workers from harm’s way — transitioning traditional equipment operators into remote site supervisors who oversee entire fleets from a safe distance.
Beyond early deployments in mining and site preparation for data centers, AIM landed a $4.9 million U.S. Air Force contract earlier this year to deploy autonomous machines for airfield repair and base construction in remote or high-risk zones. The military work builds on the company’s growing momentum following a $50 million funding round backed by Khosla Ventures, General Catalyst, and Human Capital.
AIM has risen to No. 110 on the GeekWire 200 ranking on top Pacific Northwest startups.
To support its growth, AIM has rapidly expanded its headcount, doubling in size to about 80 employees in the last few months. Sadilek is attracted to the Seattle area’s intersection of hardware expertise from companies like Boeing and Amazon alongside top-tier software and AI talent.
But while AIM has managed to hire a couple former SpaceX engineers to build out its team, it isn’t the only startup mining that rocket-engineering pedigree. TerraFirma, an Austin-based company founded by two more SpaceX engineers, raised $115 million earlier this month in the burgeoning race to semi-automate physical construction.
For Sadilek, anchoring his team in Redmond rather than Silicon Valley was a deliberate decision to stay rooted in physical engineering. Having spent years in the Bay Area during his time at Google, Sadilek wanted to avoid the tech industry’s “echo chamber.”
“I wanted to be somewhat shielded from the Kool-Aid in Silicon Valley,” he said. “We wanted to build something that’s real and gets in the black really quickly… To do that, you need to do it in an environment that is more anchored in reality.”
That philosophy extends directly into their field testing. AIM’s regional proving grounds in the mountains near Monroe, Wash., expose the autonomous equipment to heavy snow and inclement weather early in development so the physical AI is built for harsh, real-world conditions from day one.
Amid the hard hats, safety vests and construction-related decor in AIM’s headquarters space, one piece of art offers a fun take on where AIM has been and where it’s headed.
The 1949 photograph, titled “Refueling the Sunkist Lady,” shows a Jeep driving beneath a low-flying plane and transferring supplies to aid the crew during an endurance flight.
Sadilek likes it as a reminder of getting started, and what it feels like to build a company from scratch, literally working on the airplane while it’s already rolling down the runway.
“The first years of AIM were exactly like that,” he said. “I think every tech startup is like that in the early days. The problem is that some of them never finish building it before the runway ends.”
