Database startup TileDB Inc. said today it has closed on a $34 million Series B funding round, which will help it to advance its vision of becoming the “modern database” for enterprises, developers and data scientists alike.

Today’s round was notably led by AlleyCorp, a venture capital firm founded by the MongoDB database creator Kevin Ryan. Other participants in the round include Two Bear Capital, Nexus Venture Partners, Big Pi Ventures, Intel Capital, Uncorrelated, Lockheed Martin Ventures, Amgen Ventures, NTT Docomo Ventures, Verizon Ventures, S Ventures, LDV Partners and Scale Asia Ventures.

TileDB says it has been able to attract so many investors because its cloud database is uniquely capable of handling multiple modalities, or various kinds of data. At the same time, it consolidates reproducibly runnable code and versatile compute within the same product.

The database’s versatility stems from its use of a multidimensional array as its first-class data structure. According to the startup, this array can morph to effectively capture any kind of information, be it traditional data that’s stored in tables, genomic data, images, graphs, key-values, point clouds and more. It can even store vectors, which are numerical representations of unstructured data that can be especially useful for training artificial intelligence models.

In other words, TileDB can be thought of as a database jack-of-all-trades. In addition, it has the ability to run complex “extract, transform and load” or ETL pipelines, combining data from many different sources into a single, consistent data store that can be loaded into a data warehouse or another target system.

TileDB can also build data pipelines and query algorithms, and all of this happens within its serverless distributed computing environment. The startup says that by keeping data, code and compute in a single location, it’s able to eliminate data silos and increase productivity across multiple teams.

Founder and Chief Executive Stavros Papadopoulos said organizations today are faced with increasingly complex datasets and compute needs. To cope with this, they typically build complex data stacks made up of convoluted data, code and compute tools.

“This approach is extremely costly and hard to implement correctly, and cumbersome to maintain, due to the disparate tools and data engineering involved,” Papadopoulos said. “Think of TileDB as the modern data stack in a box.”

The advantage of its multimodal data stack approach is lower costs, according to TileDB. The reason is that it eliminates unnecessary software licenses, has lower cloud consumption and savings on data engineering resources, and produces faster time to insights, thanks to the speedy performance of its database.

TileDB has actually been around for some years already, first achieving prominence when it was named as one of a clutch of data-focused startups to secure funding from Intel Capital, the venture arm of Intel Corp., back in 2017. In fact, the startup was born out of a project at the Intel Science and Technology Center for Big Data. Since then, it has enjoyed traction with a number of Fortune 500 companies, as well as pharmaceutical firms, hospitals, government and defense industry organizations, the company said.

Doug Henschen, an analyst with Constellation Research Inc., agreed that TileDB is a promising all-purpose data engine platform. “In a world in which most organizations already have deep incumbent tech investments and expertise, TileDB is wisely going after new and niche use cases,” he said. “It’s looking at use cases such as genomics, LiDAR data management, earth observation and biomedical imaging to get itself established and build a case for broader adoption.”

Although it remains niche for now, Papadopoulos asserted that TileDB is also the perfect database for new generative AI models such as ChatGPT, thanks to its support for vector search. He explained that vector search is enabled by combining vector embeddings, which are natively handled by TileDB, with the machine learning models that produce the embeddings and the large language model itself. TileDB provides a way to run LLMs securely on private vector data, making it especially useful for enterprises that want to safeguard their generative AI workloads.

“Generative AI is a superb use case for the modern database,” Papadoupoulos said. “It demonstrates TileDB’s vision of a modern database is viable, valuable and future-proof.”

Image: TileDB