CrowdFlower, a San Francisco-based data enrichment, data mining and crowdsourcing platform for data science teams, last week announced its recent $10M venture funding. The investment round was led by Canvas Ventures, Trinity Ventures and Microsoft. The capital will be used to fuel adoption of CrowdFlower AI, which combines training data, machine learning and human-in-the-loop in a single platform.
"We've seen companies like Tesla and Uber build large data science teams and adopt AI and machine learning to solve billion dollar problems like driverless cars," said Lukas Biewald, founder and chief executive officer at CrowdFlower. "But we wanted to bring AI and machine learning within the reach of every business to attack million dollar problems such as classifying customer support tickets or generating customer insights from social data."
"Our team at Canvas has long been involved in AI with early investments in Siri and Nuance," said Rebecca Lynn, cofounder and partner at Canvas Ventures. "We believe corporate boardrooms are increasingly making AI a priority."
With CrowdFlower AI, customers can create machine learning models using human-labeled training data from within the platform. These models can then be deployed using humans-in-the-loop for when the model's predictions fall below a customer defined confidence threshold. CrowdFlower seeks to provide companies a single platform that can convert instructions in plain English to commercially viable machine learning models that replace humans in some part of a business process.
"At Microsoft, we're looking to create experiences for people and businesses where technology intelligently supports what they're doing," said Nagraj Kashyap, corporate vice president, Microsoft Ventures.
"CrowdFlower's approach – combining human and machine intelligence to solve all types of unstructured data problems – aligns with that effort. We look forward to supporting them in their next phase of growth in the broader machine learning and AI market."
"The industry has been focused on who has the best algorithms," said Biewald. "But when we talked with our data science customers, we saw that wasn't the problem preventing them from adopting AI more broadly. They wanted to find a way to take AI from the science lab to the board room. To do this they need both enough human labeled training data to get started, and then humans-in-the-loop to handle the cases where the model wasn't confident enough in its prediction. So we decided to solve these two problems in a single platform."