Meet Zoox, the Robo-Taxi Start-up Taking on Google and Uber

Exclusive details and images of the stealthy startup’s experimental vehicle

7 min read
Zoox self-driving car prototype
Zoox’s experimental robo-taxi in the garage at SLAC appears to have multiple lidar modules. The vehicle is conceived to have modules that drive, camber, and steer each wheel independently. So it will be able to drive in either direction.
Photo: Mark Harris

You wait for years for a self-driving taxi, then four come along at once. Google and Uber’s efforts have been well publicized, and IEEE Spectrum broke the news in February that Nissan was also developing a robotic cab. Now we can reveal exclusive details about a startup that hopes to put fully autonomous taxis on the road by 2020.

The company is called Zoox, and it’s the brainchild of the Australian designer Tim Kentley-Klay and Jesse Levinson, an engineer who worked at Stanford University with Sebastian Thrun, the first director of Google’s self-driving car program. Their vision is for a sleek, modernistic, deluxe electric taxi with gullwing doors, in which four passengers face one another. The car is code-named L4, a play on the National Highway Traffic Safety Administration’s classification of full automation as Level 4. Unlike rival designs, it has no front or rear end but can drive equally well in either direction. It has no windshields facing either way, nor does it have a steering wheel or brake pedal.

Zoox self-driving car concept The original Zoox concept. Illustration: Zoox

 Speaking at a conference in Berlin last year, Kentley-Klay said, “At the moment, mobility is crushing the soul: Don’t speed, don’t drink, don’t text. But what if we [ask], How can this stuff be awesome? What inspires me…is giving back people their lifestyles, so they can do what they want to do: texting, vegging out, drinking.” The Zoox test mule, which is currently being built in a garage in Menlo Park, Calif., is based on an innovative Swedish research vehicle with wheels that can be steered, driven, braked, and cambered independently.

Late last year, Zoox secured seed funding from a major venture capital firm and entered stealth mode to avoid tipping off rivals. It declined to comment for this story. Nevertheless, Spectrum has pieced together the tale of how a designer from Australia with no auto industry experience intends to revolutionize 21st-century transportation, including a first look at the technology Zoox is using to develop its vehicle, the venture-capital firm funding it, and exclusive photos of the car and workshop.

Kentley-Klay’s first career was in commercial media. He founded a successful animation studio, a digital media company, and a video production house, and he spent most of the past decade directing television commercials and building his businesses. Then, in 2013, he suddenly switched gears. Struck by the potential of emerging self-driving technologies, he sketched out concept designs and realized he was thinking about building the world’s first autonomous taxi.

“I know that sounds crazy,” he writes on his website. “But when you think through the mobility problems we face against the imagination of what autonomous technology allows, you would be crazy not to do it!”

Kentley-Klay formed a company called Zoox Pty. in his hometown of Melbourne and began talking with technology and car companies around the world. He visited the University of Sydney, Stanford, the MIT collaborative project in Singapore, and Singularity University, in California, to share his vision: “Zoox is not an automobile company. This is what comes after the car,” he wrote on his site. “I have in my head a concept design…so extreme it will have philistine giraffes gone wrong like [former Top Gear host] Jeremy Clarkson throwing back shots of Prozac to calm down.”

He imagines a Zoox L4 driving across a road to collect passengers without doing a three-point turn, and with airbags on the outside to mitigate crashes before they happen. “We’re even toying with this idea of ‘lights outs’ mode in the vehicle, where anything that’s being recorded stops and you could have some fun,” he said with a laugh in an interview at the Tech.AD conference in Berlin last February.

Early in 2013, Kentley-Klay got in touch with Anthony Levandowski, the engineer whose startup Google secretly bought to kick-start its self-driving car program. Levandowski invited him to Mountain View that June to hear more. In an interview with Driving the Nation at the Los Angeles Auto Show later that year, Kentley-Klay revealed, “I did a chat with his team and took a drive [in Google’s self-driving Lexus]. It was great to see they were very curious and interested in my designs.”

Zoox cofounder Tim Kentley-Klay and Google self-driving car engineer Anthony Levandowski Zoox cofounder Tim Kentley-Klay and Google self-driving car engineer Anthony Levandowski. Photo: Zoox

He might now be regretting his openness. Less than a year later, with Zoox still little more than a trademark and a Twitter account, Google unveiled its own self-driving taxi. The two-seater lacked Kentley-Klay’s design flair, but the concept was extremely similar: an urban Level 4 cab controlled by wire, with no physical driving controls.

Kentley-Klay needed to catch up fast. That meant finding engineers who could build autonomous vehicles and investors willing to spend big. Only one place had both: Silicon Valley. Kentley-Klay set his sights on Jesse Levinson, who had been working on autonomous cars at Stanford University for nearly a decade. Levinson helped write navigation algorithms for Junior, an autonomous Volkswagen Passat that took second place in 2007’s DARPA Urban Challenge and subsequently laid the groundwork for Google’s self-driving cars.

Zoox engineers Zoox designers and engineers moved into an old fire station on the grounds of Department of Energy’s Stanford Linear Accelerator Center, in Menlo Park, Calif. Because it’s federal land, the company won’t have to follow California’s robocar rules while testing there. Photos: Zoox

Levinson’s Ph.D. supervisor at Stanford was Sebastian Thrun, who left the university to head the Google project. “Jesse was one of the smartest students I ever had in my 20-year academic career,” Thrun told IEEE Spectrum. “In recent years, he was the driving force behind Stanford’s self-driving car activities.” But securing Levinson would be impossible without giving him the funds to build something.

Enter Steve Jurvetson of venture capital firm DFJ. Jurvetson is famous for having spun a US $1 million stake in Hotmail in the 1990s into hundreds of millions of dollars for his investors. He invested more than $10 million in Tesla when it was far from certain Elon Musk would deliver the Model S, and $20 million in SpaceX before the company had even flown its Falcon rocket. While DFJ would not disclose to Spectrum the amount (or even confirm the fact) of its investment in Zoox, the company briefly had a profile on recruitment website AngelList that described itself as “a DFJ-funded robotics startup.”

Venture capitalist Steve Jurveston with Kentley-Clay and others from Zoox Venture capitalist Steve Jurveston with Kentley-Clay and others from Zoox. Photo: Steve Jurvetson

In November 2014, on the company’s Facebook page, Zoox confirmed Levinson’s appointment as chief technology officer. Things were now moving fast. Zoox licensed Stanford’s patents for autonomous vehicle software and, in December, signed a lease for a disused fire station on the campus of the Department of Energy’s Stanford Linear Accelerator Center, in Menlo Park. The company moved in a couple of weeks later. A major benefit to working on a closed campus is that Zoox does not need to apply for an autonomous vehicle testing permit from the state or abide by the associated restrictions.

Kentley-Klay is characteristically enthusiastic about the workshop, calling it in a press release “the startup garage to end all startup garages.” Publicly accessible visa documents reveal that Zoox also went on a hiring spree, recruiting a director of strategy and a media manager from Australia, as well as roboticists, machine vision experts, and software engineers from the Valley.

But building a car from scratch is no mean feat. Big carmakers can take a decade to design new platforms, and even Google and Tesla started out by adapting existing vehicles. Luckily, research happening half the world away was about to make Zoox’s birth much easier.

For the previous dozen years, KTH Royal Institute of Technology, in Stockholm, had been working with autonomous corner modules, a Volvo idea that packs almost everything needed to drive, steer, and brake an electric car into the car’s wheels. This reduces energy loss in the power train and, by tilting each wheel as required, lowers rolling and cornering resistance. Also, by including motors and brakes in every wheel assembly, the system provides redundancy, improving safety. The resulting modules are heavier than normal wheels, however, and require sophisticated control systems to perform at their best.

In 2012, KTH began building a research concept vehicle (RCV), a bare-bones two-seater with four autonomous cornering modules. The 400-kilogram car has battery power for 30 minutes of driving and a top speed of 70 kilometers per hour (45 miles per hour). Because its driving functions are controlled by wire (rather than drive shafts) and its control systems are modular, it makes an excellent autonomous test platform. Researchers from KTH’s Integrated Transport Research Lab met Zoox at a conference and explained their concept. It was a perfect match for what Zoox was planning.

Kentley-Klay imagines the L4 as not only bidirectional but identical at each corner, with four independent motors, batteries, and computers. “This means we only need to design and test 25 percent of the components to make 100 percent of the car,” he said at the L.A. Auto Show. With DFJ’s money, Zoox bought the first—and so far, only—RCV in private hands. (Although KTH would not confirm a purchase price, the RCV cost around $200,000 to develop.)

Zoox quickly got to work on it. On an unannounced visit to the company’s garage in April, the RCV was almost unrecognizable. A roof above the roll cage sprouts a GPS antenna, and there seem to be  Velodyne lidar modules on each corner. There are more lidar sensors  above the wheels and a forward-facing array that could be sonar sensors.

If Kentley-Klay can stick to a timeline he shared with ReadWriteDrive in 2013, Zoox will be aiming to finish a drivable prototype by the end of next year, adapt it for public roads by 2019, and debut L4 taxis in a sunny city like Las Vegas as early as 2020. Between now and then, Zoox will have to navigate a maze of regulatory, testing, and manufacturing challenges—unless it is simply bought by Uber or Google first.

With some of the world’s biggest technology and car companies already testing their vehicles in public, Kentley-Klay has seemed realistic about Zoox’s chances. “There’s no middle ground,” he said last year in Berlin. “We will either succeed very, very well or fail extremely well. It’s an all-or-nothing proposition.”

About the Author

Washington-based Mark Harris has been investigating the origins of self-driving cars and how they’re regulated.

The Conversation (0)

Self-Driving Microscopes to Navigate the Nanoscale

Deep learning as the ultimate tour guide

4 min read
A man in safety glasses in front of microscopy equipment

Maxim Ziatdinov of Oak Ridge National Lab in front of a scanning tunneling microscope.

Maxim Ziatdinov

It's difficult to find an area of scientific research where deep learning isn't discussed as the next big thing. Claims abound: deep learning will spot cancers; it will unravel complex protein structures; it will reveal new exoplanets in previously-analyzed data; it will even discover a theory of everything. Knowing what's real and what's just hype isn't always easy.

One promising—perhaps even overlooked—area of research in which deep learning is likely to make its mark is microscopy. In spite of new discoveries, the underlying workflow of techniques such as scanning probe microscopy (SPM) and scanning transmission electron microscopy (STEM) has remained largely unchanged for decades. Skilled human operators must painstakingly set up, observe, and analyze samples. Deep learning has the potential to not only automate many of the tedious tasks, but also dramatically speed up the analysis time by homing in on microscopic features of interest.

Keep Reading ↓ Show less

100 Years Ago RCA Sent the First Voice Message Over the Air

It was transmitted through Radio Cental’s relay network

4 min read
A photo of people sitting as at a table with people walking around behind them.

Operators working at the RCA Radio Central transmitting facility, in Rocky Point on New York’s Long Island.


When the Radio Corporation of America was founded in 1919, one of its missions was to become the world’s leading radio communications company. Among RCA’s first steps was to build a center that would transmit international signals and telegraphic messages. In 1921 the company opened Radio Central, in Rocky Point on New York’s Long Island.

The transmitting facility was connected to a radiogram-receiving facility in Riverhead—about 24 kilometers away on Long Island—through a system of relays using existing telephone lines. The two locations worked with the Central Traffic Office in New York City to send and receive messages.

Keep Reading ↓ Show less

EP29LPSP: Applications in Plasma Physics, Astronomy, and Highway Engineering

Ideal for demanding cryogenic environments, two-part EP29LPSP can withstand temperatures as low as 4K

3 min read

Since its introduction in 1978, Master Bond EP29LPSP has been the epoxy compound of choice in a variety of challenging applications. Ideal for demanding cryogenic environments, two-part EP29LPSP can withstand temperatures as low as 4K and can resist cryogenic shock when, for instance, it is cooled from room temperature to cryogenic temperatures within a 5-10 minute window. Optically clear EP29LPSP has superior physical strength, electrical insulation, and chemical resistance properties. It also meets NASA low outgassing requirements and exhibits a low exotherm during cure. This low viscosity compound is easy to apply and bonds well to metals, glass, ceramics, and many different plastics. Curable at room temperature, EP29LPSP attains its best results when cured at 130-165°F for 6-8 hours.

In over a dozen published research articles, patents, and manufacturers' specifications, scientists and engineers have identified EP29LPSP for use in their applications due to its unparalleled performance in one or more areas. Table 1 highlights several commercial and research applications that use Master Bond EP29LPSP. Table 2 summarizes several patents that reference EP29LPSP. Following each table are brief descriptions of the role Master Bond EP29LPSP plays in each application or invention.

Keep Reading ↓ Show less