AI data labeling startup Handshake has acquired the data label auditing startup Cleanlab. Handshake originally launched in 2013 as a platform for hiring recent college graduates. Approximately one year ago, the company expanded to launch a human data labeling business focused on serving foundational AI model companies. Cleanlab was founded in 2021 and provides software designed to improve the quality of data produced by human labelers.
The primary purpose of the deal is an acqui-hire to acquire talent. The transaction adds nine key Cleanlab employees to Handshake’s research organization. This group includes the startup’s three co-founders, all of whom earned computer science PhDs from MIT: Curtis Northcutt, Jonas Mueller, and Anish Athalya. The specific financial terms of the deal were not disclosed.
Cleanlab had raised a total of thirty million dollars from investors, including Menlo Ventures, TQ Ventures, Bain Capital Ventures, and Databricks Ventures. At its peak, the startup employed more than thirty people.
The researchers from Cleanlab are experts in developing algorithms that can automatically flag incorrect data without requiring a second human reviewer. This expertise will be used to improve the overall quality of the data Handshake produces for AI laboratories.
Handshake’s chief strategy and innovation officer, Sahil Bhaiwala, explained that the company has an in-house research team focused on identifying model weaknesses and determining what high-quality data should look like. He noted that the Cleanlab team has been concentrating on this specific problem for years.
Curtis Northcutt, the CEO of Cleanlab who pioneered its automated data label auditing technology, stated that the company received acquisition interest from other AI data labeling firms. However, Cleanlab chose to sell to Handshake. Northcutt explained that competing data labeling companies, such as Mercor, Surge, and Scale AI, frequently use Handshake’s platform to source specialized human experts like doctors, lawyers, and scientists for their own data labeling projects. He reasoned that if you are going to pick a company to join, you should pick the source rather than the middleman.
Handshake was last valued at three point three billion dollars in 2022. The company was forecasted to end 2025 with a three hundred million dollar annualized revenue run rate and is reportedly on track to reach a run rate in the high hundreds of millions this year. Handshake has provided data for eight top AI labs, including OpenAI.