Rockset Inc., the startup behind the speedy database of the same name, said today it’s expanding the vector search capabilities of its platform.

The new features are intended to enable more rapid search for cloud-based data to support artificial intelligence applications. Vector search has emerged as an essential ingredient for databases that power generative AI and other kinds of AI applications. It allows for unstructured data such as images and written handwriting to be stored as vector embeddings, which are mathematical structures that enable that information to be indexed and searched.

As a result, vector search enables the creation of much more powerful and accurate AI models. With the addition of vector search earlier this year, Rockset became a viable database option for generative AI models.

With today’s update, the Rockset database now supports “approximate nearest neighbor” or ANN search, meaning it can achieve “billion-scale similarity search” in the cloud. But as the company explained, having vector search is not enough. Because the most powerful large language models generate vector embeddings with many thousands of dimensions, exact nearest neighbor or ENN search becomes an extremely complex and computationally intensive task.

By adding support for ANN instead, Rockset makes it possible to create vector embeddings for any AI model and index them for fast similarity search at massive scale, the company said. The new capability is coupled with Rockset’s LlamaIndex and Langchain integrations to help developers iterate more quickly and create more relevant AI experiences, the company said.

According to the company, developers can now store and index billions of vectors alongside hundreds of terabytes of essential data such as text, JSON, geo- and time-series data. They’ll be able to build AI applications that can update themselves in real time by inserting, updating and deleting vectors and metadata stored on the Rockset database. New data is reflected in searches of the database in milliseconds, meaning it becomes available to AI models the moment it’s added.

Rockset co-founder and Chief Executive Venkat Venkataramani (pictured) said it’s not an easy feat to incorporate real-time signals and updates into vector search. Yet, AI applications won’t be nearly as useful without them. “We’ve spent years designing Rockset for real-time updates and are thrilled that companies can now build scalable AI applications on real-time, streaming data,” he added.

Rockset’s database is designed to support real-time applications of every shape, not just AI. For instance, it’s widely employed for “internet of things” devices and sensors that must record and analyze data as it’s ingested to ensure machines operate efficiently and without downtime.

Should a sensor collect data that suggests a machine is about to fail, Rockset can register this and provide feedback for instant troubleshooting to prevent it from happening. It also has applications in cybersecurity and e-commerce, where real-time data is extremely valuable.

Rockset’s speedy database can ingest and analyze data in a fraction of a second. In addition, the database is said to be more reliable thanks to its novel compute-compute separation feature, which provides two separate pools of compute capacity for ingesting records and analyzing them, as opposed to other databases that use just one pool.

Should one of the modules for ingesting or analyzing data require more processing resources than usual, it will not affect the performance of the other. According to Rockset, this can be especially useful for demanding AI applications.

The company’s claims are not unsubstantiated, as it recently closed on a $44 million round of funding. At the time, its backers said they were especially impressed with its usefulness for AI applications. Rockset also disclosed that its revenue and customer base have more than doubled in each of the last two years.

A number of Rockset’s customers are already taking advantage of its advanced vector search to deploy AI apps at scale, such as the low-cost airline JetBlue Airways Corp.

JetBlue’s senior manager of Data Science and Analytics said in a recent case study that iteration and speed of AI are the most essential factors for a database. “We saw the immense power of real-time analytics and AI to transform JetBlue’s real-time decision augmentation and automation since stitching together 3-4 database solutions would have slowed down application development,” he said. “With Rockset, we found a database that could keep up with the fast pace of innovation at JetBlue.”

Venkataramani appeared on theCUBE, SiliconANGLE Media’s mobile livestreaming studio, during its coverage of the AWS re:Invent 2022 Global Startup Program, where he discussed Rockset’s real-time analytics capabilities and the hundreds of applications it can support in more depth:

Photo: SiliconANGLE