Materialize, a company developing a streaming structured query language (SQL) database platform, today announced that it raised $60 million in series C funding, bringing the companyâ€™s total raised to more than $100 million. Redpoint Ventures contributed the capital with participation from Kleiner Perkins, Lightspeed Venture Partners, and others, and cofounder and CEO Arjun Narayan says that itâ€™ll be used to grow Materializeâ€™s engineering team and bring its cloud service from beta to general availability.
Real-time data analytics can benefit companies across finance, retail, ecommerce, and other industries. For example, banks can identify fraudulent transactions while minimizing false positives, and ecommerce sites can provide better personalization via recommendations. But real-time data analytics often requires compromises between cost, speed, and features. For example, itâ€™s difficult to achieve millisecond response times for queries without building custom microservices.
Founded in 2019, Materialize â€” whose team includes early employees of Dropbox, Stripe, and YouTube â€” offers a standard SQL interface for streaming data so that companies can build queries without the need for engineers with specialized skills. The platform computes and incrementally maintains data as itâ€™s generated, so that query results are accessible the moment that theyâ€™re needed.
â€œFrank McSherry and I founded Materialize in February 2019 after realizing the implications of his timely and differential dataflow research in providing â€˜trueâ€™ real-time data streaming. Weâ€™ve been studying this problem for decades, and Frank in particular spent years doing the hard science that allows developers to write complex queries for streaming data using standard SQL,â€ Narayan told VentureBeat via email. â€œWe have the mentality that all businesses should have access to the power of accurate streaming data without tradeoffs. Although other data streaming solutions have been around for years, each one of them requires some sort of compromise.â€
McSherry and Narayan named Materialize after the database concept of â€œmaterialized views.â€ In databases, materializing views refers to the act of precomputing the results for a query so that theyâ€™re instantly available when needed â€” rather than doing the work on-demand and waiting for the computation to finish.
â€œMaterialized views that are always fresh have long been prohibitively expensive in traditional database systems, and Materialize makes them cheap and always-ready on all of a companyâ€™s streaming data,â€ Narayan said. â€œWeâ€™ve seen our early customers use Materialize for real-time data visualization, financial modeling, and to advance various software-as-a-service applications in marketing tech, logistics, and enterprise resource planning.â€
While Materialize isnâ€™t an engine for machine learning or AI itself, Narayan notes that it can play a role in the data pipelines that feed into machine learning models. Some companies, including Datalot, have investigated using Materialize as a â€œstreaming feature store,â€ a class of tool used to store commonly used features in models.
Image Credit: Materialize
â€œCurrent solutions offer a linear tradeoff between speed and cost. If you want to move more quickly, you simply have to pay for it,â€ Narayan said. â€œWe look to break this pattern by offering extremely low latency computation, but on a much more efficient scale through standard SQL.â€
Materialize says that in six months, itâ€™s grown its developer community to over 970 people and attracted brands including Density and Kepler Cheuvreux. This month, the startup, which has close to 60 employees, plans to open its headquarters in Slackâ€™s previous New York City office.