Labelbox today announced the close of a $25 million series B round to grow its platform that helps customers label the data needed to train AI systems.
The round was led by Andreessen Horowitz, with participation from Google’s AI-focused Gradient Ventures fund, Kleiner Perkins, and First Round Capital.
The funds will be used to develop and accelerate Labelbox’s roadmap for machine learning and computer vision models by doubling the size of its engineering and sales teams.
Labelbox also enables users to automate some labeling so a company can manually label all data except any that falls below a particular prediction confidence threshold, COO Brian Rieger told VentureBeat in a phone interview.
The funding will also be used to codify best practices and standard metrics for model performance among data scientists, developers, and data engineers, in part by working with university and business partners.
“What happens today very often is that folks come out of the academic institutions, and they’ve kind of got that academic side of machine learning, but they haven’t experienced the process of taking a production system from nothing into production. And there are some common technologies, common formulas that need to be developed and understood amongst the community,” he said.
Among the standardization policies Labelbox seeks: Common data exchange file formats and the need for roles within organizations — like data-labeling operations manager — to accelerate AI deployment and advance company business goals.
“Labeling operations manager is this role that’s never been defined before globally but exists within many of the companies we work with,” Rieger said.
Labelbox has now raised $39 million to date, including a $10 million series A in April 2019. Andreessen Horowitz general partner Peter Levine will join the board as part of the latest round.
The company was founded in 2018 and is based in San Francisco, with 30 employees.