SAN FRANCISCO — Labelbox, a training data platform for enterprise machine learning applications, has closed a $110 million Series D funding led by SoftBank’s Vision Fund 2. Snowpoint Ventures and Databricks Ventures also participated along with previous investors B Capital Group, Andreessen Horowitz and Catherine Wood, CEO and founder of ARK Invest. To date, Labelbox has raised $189 million in venture funding.
“Labelbox has become a complete AI training data platform for enterprises,” said Manu Sharma, co-founder and CEO. “Our customers use Labelbox as their data engine, leveraging active learning and facilitating human supervision to relentlessly improve AI model performance.”
Labelbox’s software platform is designed to facilitate the entire training data iteration loop that improves ML model performance. It integrates a collection of tools to annotate data and train AI models, conduct error analysis to identify data on which the model performs poorly, refine annotations found to be incorrect or ambiguous, supplement data through augmentation or additional data collection and then test the model and repeat the error analysis in a continuous loop that improves model performance.
“It’s not just about annotation,” said Brian Rieger, Labelbox co-founder and President. “We cover this entire iteration loop on a single platform, continually optimizing the data with a focus on getting more and more efficient over time.”
To build real-world applications, machine learning teams need robust infrastructure that can easily import raw data into labeling workflows, allow enterprises to manage widely distributed annotation teams, transparently monitor quality, adjust for bias, and export high-quality training data to machine learning models. To deploy accurate models and drive optimal business outcomes, training data must be constantly improved. Additionally, Labelbox offers Boost – a service that features a world-class workforce and dedicated labeling expertise to ensure customers find success quickly on the platform and then continually become more efficient and effective.
“The investment in Labelbox – the first as Databricks Ventures – felt like a natural fit given the strong existing partnership between our two companies,” said Andrew Ferguson, Head of Databricks Ventures. “We started Databricks Ventures to support companies extending the lakehouse ecosystem and Labelbox’s collaborative training data platform allows companies to quickly produce structured data from unstructured data, and train AI on unstructured data on the Databricks Lakehouse. With this investment, we are looking forward to supporting Labelbox and our rapidly growing number of joint customers with streamlined, powerful capabilities.”
Labelbox automates the process with a web-based platform that pre-labels data and allows enterprises to collaborate easily across databases, BPOs and labeling services regardless of time zone or geography. Labelbox customers report accelerating iteration cycles by up to 800 percent using the platform and cutting in half the time it takes to push new models into production.
“Data is the new oil and labelling is one of the most essential parts of the refinery,” said Robert Kaplan, Investment Director at SoftBank Investment Advisers. “We believe that Labelbox has the most advanced end-to-end training data platform focused on collaboration, automation and data quality that simplifies the time-intensive process of data labelling, allowing technical resources to focus on performance and getting AI to production faster. We are delighted to partner with Manu Sharma and the team to support their mission to democratize access to AI development.”