AI data labeler Handshake buys Cleanlab, targets several other acquisitions

AI data labeling startup Handshake has acquired data labeling audit startup Cleanlab, the company told TechCrunch.

Handshake started in 2013 as a platform to hire college graduates and launched a human data labeling business about a year ago to serve basic AI modeling companies. Cleanlab, founded in 2021, is a startup that provides software to improve the quality of data generated by human labelers.

The purpose of the deal is primarily to acquire talent, aka acqui-hire, adding nine key Cleanlab employees to the Handshake research organization. It includes the startup’s founders, who earned PhDs in computer science from MIT: Curtis Northcutt (pictured above), Jonas Mueller, and Anish Athalye. Terms of the transaction were not disclosed (although, as previously reported, sometimes acqui-hire can be surprisingly lucrative for the founder.)

Cleanlab has raised a total of $30 million from investors, including Menlo Ventures, TQ Ventures, Bain Capital Ventures, and Databricks Ventures. At its peak, the startup had more than 30 employees.

Cleanlab researchers are experts in developing algorithms that flag erroneous data without other human researchers. The goal is to improve the quality of the data Handshake generates for the AI ​​lab.

“We have an in-house research team that thinks a lot about where our model is weak, what we should be producing? How high quality is that data?” Sahil Bhaiwala, chief strategy and innovation officer at Handshake told TechCrunch. “The Cleanlab team has been focusing on this problem for years.”

Northcutt, Cleanlab’s CEO who is credited with pioneering automated data label auditing, said the company has received acquisition interest from other AI data label companies. But the startup chose to sell to Handshake because, it says, its data labeling competitors, including Mercor, Surge, and Scale AI, often use Handshake’s platform to source human experts such as doctors, lawyers, and scientists for data labeling projects.

“If you’re going to choose one, you should choose the source, not the middleman,” Northcutt told TechCrunch.

Handshake, which will eventually be worth $3.3 billion by 2022, is projected to end 2025 with an annualized revenue rate (ARR) of $300 million and the news on track to achieve ARR of “high hundreds of millions” this year. The company provides data to eight top AI labs, including OpenAI.