As machine learning becomes a more integral part of running businesses, the model-building process still requires iteration and experimentation. Comet has created an entire platform to get models from idea to product, and today the company announced a $50 million Series B.
The investment comes on the heels of the company’s $13 million A round in April.
OpenView led the round with participation from existing investors Scale Venture Partners, Trilogy Equity Partners and Two Sigma Ventures. The company has now raised almost $70 million, according to Crunchbase data.
Company co-founder and CEO Gideon Mendels said that the product works on any platform from a laptop to the cloud to an on-prem cluster.
“Comet is a platform that empowers data scientists and machine learning engineers to accelerate machine learning development, and the way we do that is by managing and optimizing the entire machine learning lifecycle, so all the way from experiment tracking to model production monitoring,” Mendels told me.
He said that the approach has been working: The company’s annual recurring revenue (ARR) grew 5x this year, and Comet now counts 150 businesses using the platform, including Uber, Zappos and Etsy.
Mackey Craven, partner at lead investor OpenView, said that he was attracted to Comet because he sees it as a startup building a viable product for an emerging market with a big opportunity.
“When we see that rare combination of an exceptional founding team that can be the core of a large and enduring market opportunity that is large enough to support them and then transition in that market, either because it’s the creation of a new market or [because of] technological dislocation that allows a new entrant to create and capture the value in that market going forward, we step up and we step up big,” Craven explained.
Today, the company has 50 employees in nine different countries on four continents with plans to reach 100 by next year. Mendels said that diversity and inclusion are a key part of the company’s value system.
“It’s just really a key part of our culture,” he said. “We have 35% of the team from underrepresented groups today and that continues to be a focus for future hires.”
The company is also launching a new product called Artifacts, a data versioning tool that works in the same way as a document versioning function, letting data scientists track the ways that the data has changed over time.
“With Comet Artifacts, essentially when you’re working on a machine learning pipeline, we automatically version every snapshot of the data. So every time you make a change, we make a version of it,” Mendels said, adding that this offers a number of advantages, primarily enabling data scientists to see how the data they used to train the model changes over time and compare the training model data to what’s in the model in production.