Supplier to Self-Driving Car Makers Nears Financing at $1 Billion Valuation

Scale launched in 2016 after receiving funding from Y Combinator, the startup accelerator. The 22-year-old Scale CEO Alexandr Wang, who co-founded the company when he was 19 after deferring college admission in order to work at Quora and then dropping out of the Massachusetts Institute of Technology, didn’t have a comment for this article. 

Last year Scale, which has nearly 100 employees, raised $18 million, according to PitchBook. Among its customers are Alphabet’s Waymo, Uber, Zoox, Voyage, General Motors’ Cruise, Nuro and Lyft. Some of them spend more than $1 million a year with Scale, according to people at self-driving vehicle developers that use its services. 

Scale’s broader ambition is to help machine learning researchers train their computer vision algorithms for a variety of uses. It is already working with teams in robotics and drones, along with retail—to enable technologies such as those behind Amazon’s Go stores, which uses cameras and AI to track the items customers put in their shopping baskets.

Scale and its rivals currently use a crude method to put digital labels on imagery and other data to improve the visual recognition systems in self-driving cars. They find low-wage laborers in other parts of the world to view images and other computerized representations of the physical world collected by sensors on clients’ prototype vehicles. The workers then manually identify objects like cars, pedestrians and cyclists. 

With enough labeling, the idea goes, the clients’ algorithms will get better over time at automatically identifying such objects on their own. Eventually, data annotation firms like Scale believe they will be able to automate the data-labeling for their customers, without relying on humans.

One reason Scale broke out from the pack was because it was among the first firms to annotate three-dimensional data that self-driving vehicle developers collected using lidar and radar sensors, after starting with two-dimensional camera-based images. After it became clear that Scale was the runaway market leader, one of the company’s chief rivals, Mighty AI, recently sold itself to Uber.

Startup valuations have dropped in many parts of the self-driving vehicle industry as timelines shift further into the future for when driverless vehicles will transport people and freight. Money still continues to flow into the category though. That includes funding for vendors that sell software to well-funded companies like Waymo and Cruise that are working on ride-hailing services that use self-driving vehicles. 

Investors also continue to back makers of software for simulating the performance of autonomous vehicles on the road and tools for remotely connecting to prototype vehicles to help them when they get stuck, along with developers of lidars and other sensors for self-driving vehicle prototypes.