FeatureByte, the developer of a feature engineering platform, launched from stealth with a $5.7 million seed round. The Boston-based startup bills its SaaS solution as built specifically for data scientists to simplify the creation, operation, management, and monitoring of machine learning functions.
A function, also known as a variable, is any measurable input used for a predictive machine learning model. As the first step in the development of ML models, feature engineering is the process of applying domain knowledge (for example, business knowledge, mathematics, and statistics) to extract analytic representations from raw data. To choose the most useful features for a given model, data scientists must select, manipulate, and transform raw data into a format that can be directly used by ML models. There are plenty of data preparation tools created for the data analytics domain that can automate data preparation, but there is a lack of automation tools built specifically for AI model workloads.
Given this limitation of data preparation, feature engineering and management is a complicated process that can be slow and expensive. According to Gartner, functions are among the most managed and refined data assets because of the amount of time, effort, and skill involved. FeatureByte says that despite this importance, many organizations do not have an effective feature management system. In addition, the company says feature engineering is the weakest link in scaling AI, as it “requires the confluence of three unique skills: domain knowledge, data science, and data engineering. Even in organizations with mature AI practices, these areas of expertise live in silos. And there is a lot of friction at the intersection of these silos.”
Solving the problem of these disparate expertise silos is the goal of FeatureByte co-founders Razi Raziuddin and Xavier Conort, both DataRobot alumni. FeatureByte CEO Raziuddin scaled DataRobot from 10 to 850 employees and led its go-to-market strategy. Conort, CPO at FeatureByte, was chief data scientist at DataRobot and built his R&D data science team. The $5.7 million starting round was led by Glasswing Ventures and Tola Capital, and the company plans to use the money to scale its R&D and go-to-market operations.
“Our team has successfully launched AI implementations for hundreds of organizations worldwide. The one constant challenge that enterprises face is feature engineering and management. Xavier and I founded FeatureByte to radically simplify the process for data scientists and application developers,” said Raziuddin. “The market is extremely fragmented, with silo solutions only addressing pieces of the puzzle. We are developing a solution from first principles to address the full engineering cycle and are excited to partner with Glasswing Ventures and Tola Capital to advance this vision and mission.
FeatureByte has plans for its cloud-based platform to have integration capabilities with Snowflake and Databricks. FeatureByte will be available to early users through an invite-only beta program. Visit this link for more information.
Feature Stores emerge as indispensable technology for machine learning
What is Feature Engineering and why should it be automated?
Artificial intelligence and machine learning threaten to become a major bottleneck — here’s how to solve it