Stealthy Self-Driving Startup Ghost Locomotion Raises $15 Million, Filings Show

The SF Bay Area is home to a lot of self-driving car initiatives.

Some startups, like Zoox, are trying “to make the whole widget” by designing a car and its tech from bumper to bumper.

Follow Crunchbase News on Twitter

Ghost Locomotion, a rather stealthy startup based out of Mountain View, appears to be taking a different approach. It is designing and developing core self-driving tech which can be used to retrofit existing cars or enhance new ones. In this way, Ghost is similar to Comma.ai, another company in the space which Crunchbase News covered before.

Today, the company filed paperwork with the SEC indicating it has raised $15 million in outside funding.

Who’s Involved

The filing lists the company’s two co-founders: CEO John Hayes and CTO Dr. Volkmar Uhlig.

Prior to Ghost Locomotion, Hayes co-founded Pure Storage (which went public in 2015) and served a stint in the office of Yahoo’s CTO. Uhlig earned a PhD in computer science at Karlsruhe Institute of Technology in Germany before starting his career at IBM Watson Research. After IBM, but before Ghost Locomotion, he was the CTO of Adello and a systems architect at Teza Technologies.

The filing also lists a number of non-executive directors who have probably invested in the company.

These likely investors include:

• Keith Rabois, a partner at Khosla Ventures who made a prior seed-stage investment in Lyft back in 2010. Khosla’s recent automotive investments including dashcam maker Owl Cameras and autonomous transport fleet manager Voyage.
• Gene Berdichevsky, the seventh employee at Tesla who left to start Sila Technologies, which aims to improve lithium-ion batteries.
• Michael Speiser, a managing director at Sutter Hill Ventures who led his firm’s investments in Pure Storage, John Hayes’s prior startup.

There’s reason to believe Ghost Locomotion has some prior funding, but it’s difficult to tell how much. We were unable to find historical regulatory filings by the company.

However, it appears as though the company was initially incubated out of Sutter Hill Ventures. Ghost Locomotion’s CEO did a six-month stint as an entrepreneur in residence at the venture firm immediately prior to starting the company.

At time of writing, the company’s LinkedIn page lists 29 employees. Considering that Bay Area software engineering salaries, specifically specialists in self-driving, are very high, the company has presumably been able to access plenty of capital to date.

No prior relationship between Khosla Ventures and Ghost Locomotion could be found, leading us to conclude that Khosla Ventures is a new investor in the company.

What The Company Is Building

From the company’s website, it’s difficult to tell what, exactly, it is building.

It suggests that Ghost’s system involves “a few discreet cameras.” Beyond that, hardware details are scant. But the experience its marketing copy portrays is clear.

“Ghost is a driver, designed to drive like you,” says one snippet. “Ghost transforms your commute, freeing your hands from the wheel and your mind from the road,” says another.

One of the biggest hurdles to building autonomous vehicles is obtaining a corpus of data to train the statistical models which, in turn, govern vehicle behavior.

A company like Google, by virtue of its mapping efforts, was well-positioned to launch a self-driving vehicle division.1 It already had the data.

So what’s a startup to do? Find a way to gather that data itself.

Ghost Locomotion’s website says its technology is “Trained by observing millions of real-world miles driven by real people.”

Where is it obtaining that data? In April, Ghost Locomotion launched a dashboard camera app, called MileMarker, on the Google Play Store. The app serves as a dashcam, but it also logs users’ GPS locations to let them track mileage and log trips. It offers unlimited cloud storage to its users.

The app’s privacy policy says it “collects and analyzes video and driving sensor recordings to provide an enriched presentation to [the user]” and will store any “source data […] in an anonymized fashion.”

From just the time-stamped location data, it’s possible to derive average speed, how fast traffic moves in certain segments of road, and how that changes throughout the day. And that doesn’t account for what can be found in video or sensor data.

As far as strategies for bootstrapping a driving dataset go, launching a dashcam app is tried and true.

That’s what Comma.ai, another self-driving car company, did. In our coverage of a $5 million fundraise by Comma, we noted that it also launched a dashcam app, presumably to gather similar types of data.

As far as Ghost Locomotion’s “driver” goes though, it’s unclear when it will ship or what its final form is. Early adopter-types can reserve theirs today.

Self-driving car technology is one of those fields that constantly feels “just a few years away” from maturity. All these companies need to do is develop a system that’s as safe as a human behind the wheel. It’s not rocket science!

It’s easier said than done, of course, but that doesn’t stop entrepreneurs from trying.

Illustration: Li-Anne Dias