OpenAI purchases enterprise data search startup Rockset

OpenAI, the company behind ChatGPT, today acquired Rockset Inc., a startup known for a high-speed database of the same name designed for information retrieval.

The aim is to provide the artificial intelligence company a platform for its customers to better index and query their data. Rockset’s platform enables users, developers and enterprise businesses to rapidly search capabilities of vast amounts of data, which is a fundamental need for AI applications.

OpenAI said in its announcement that it plans to integrate Rockset’s technology into its platform to allow its customers to build better, more accurate and efficient AI applications. “Rockset’s infrastructure empowers companies to transform their data into actionable intelligence,” said Brad Lightcap, chief operating officer of OpenAI.

Although primarily focused on delivering AI large language models and related services, OpenAI also has to contend with the needs of businesses that include data retrieval and augmentation. It’s also in competition with rivals such as Google LLC, which has access to vast cloud storage, retrieval and compute capabilities that it can offer its customers. This could make the acquisition deal with Rockset beneficial for OpenAI’s customers by providing a highly responsive cloud database for AI information retrieval.

“Rapid advancements in LLMs are enabling a Cambrian explosion and numerous innovations across every industry, driving a preponderance of AI applications,” said Rocket Chief Executive Venkat Venkataramani (pictured). “While the nature of these applications has changed, the underlying infrastructure challenges have not. Advanced retrieval infrastructure like Rockset will make AI apps more powerful and useful.”

Rockset raised $44 million in August led by Icon Ventures, with participation from Sequoia Capital and Greylock Partners, raising the total invested in the company to $144 million.

In late 2023 the company enhanced its vector-search capabilities to enable AI applications at scale, which provide the capability to add “embeddings,” or multidimensional numerical representations of text, images and other objects for search. Vector embeddings make it possible for AI and machine learning algorithms to understand the contextual relationships between words or other objects, such as how “dogs” and “cats” are both animals, which makes conversational LLMs or recommendation engines more accurate. The company quickly combined this into a native hybrid offering that will allow AI apps to do metadata keyword searches and vector searches in the same query, saving time.

Venkatarimani added that the acquisition will allow AI to be more accessible to developers and enterprise businesses in “a safe and beneficial way.”

The CEO spoke to theCUBE, SiliconANGLE Media’s livestreaming studio, in 2022 about the need for a real-time analytics database:

Photo: SiliconANGLE