Five-year-old startup Scale AI Inc. is worth $7.3 billion after a new $325 million funding round, announced Tuesday, that was jointly led by Dragoneer, Greenoaks Capital and Tiger Global.

The startup revealed on the occasion that Jeff Wilke, the former head of Amazon.com Inc.’s worldwide consumer business, is joining as an adviser to Chief Executive Officer Alexandr Wang.

Scale AI provides customers such as OpenAI with training data they can use to build their artificial intelligence software. Developing an AI involves creating a neural network and then training that neural network on a collection of sample information similar to the info it will analyze in production. Through repeated trial and error, the AI learns ways of producing results with higher accuracy and using less processing power.

Many of the most cutting-edge machine learning algorithms are available under a free open-source license. As a result, assembling the training dataset is often the most difficult part of AI development, particularly in large projects where developers might need millions of individual data points.

Scale AI eases the process for its customers. A company can turn to the startup with a raw training dataset, such as road measurements taken using lidar sensors, and Scale AI will add annotations, the missing element necessary to make the data useful for AI training. Annotations in a machine learning context are descriptive labels attached to a training file that highlight objects of interest, such as cars captured in a lidar scan. The labels guide how the AI learns to identify patterns. 

Scale AI generates annotations with the help of human experts, who manually annotate customers’ raw datasets, and neural networks that automate certain parts of the task. The prepared data is then reviewed for mistakes by AI quality assurance systems. For added measure, Scale AI says, its human experts also take part in the review process when needed to catch subtle errors that otherwise might go unnoticed.

In addition to OpenAI, the startup’s customer list includes Nvidia Corp., General Motors Co. and SAP SE, along with a long list of other big enterprises. With $325 million in additional funding on its books, Scale AI should be well-positioned to keep growing that installed base even amid the growing competition from other AI training dataset startups.

The startup is looking beyond data annotation as part of its expansion plans. Last August, Scale AI introduced Nucleus, a software tool that helps developers find ways of improving their training datasets. A self-driving car startup, for example, could use Nucleus to determine whether adding more lidar scans of highways to its dataset could improve the reliability of its autonomous driving system.

The rapid adoption of machine learning in the enterprise is lifting demand for both training data and AI software tooling. Scale AI is positioned to address both trends, which, judging by its latest $7.3 billion valuation, gives it significant growth opportunities.

Dragoneer, Greenoaks Capital and Tiger Global were joined by several other backers in the round. They included Wellington Management, Durable Capital, Coatue, Index, Founders Fund and Y Combinator. 

Image: Scale AI