Well-funded vector database startup Pinecone Systems Inc. today announced a serverless version of its product aimed at artificial intelligence applications.

The company said the serverless architecture, which is a cloud computing execution model where the cloud provider dynamically manages the allocation and provisioning of servers, can reduce costs by up to 98%. It can also eliminate the need for developers to provision infrastructure, allowing applications to be brought to market faster.

Pinecone cited its own research that showed that making more data available for context retrieval reduces the frequency of unhelpful answers – or those that incorrectly respond or fail to respond to a question — by 50%. It said large language model developers can improve the quality of generative AI models simply by making more data accessible.

Unlike relational databases, which store data in rows and columns, vector databases represent unstructured data as high-dimensional data points, each representing a vector or an array of numbers. One of the primary functions of a vector database is to perform similarity searches which can quickly find vectors that are most similar to a given query vector using measures like cosine similarity or Euclidean distance.

Storing and searching through large amounts of vector data on-demand can be prohibitively expensive, even with a vector database, the company said. Pinecone’s technology stores and searches AI-generated representations of unstructured data that encapsulate the meaning of the original content in a machine-readable format, making it more efficient for keyword-based searches. The serverless architecture eliminates the infrastructure provisioning stage to save time and money.

In addition, Pinecone’s architecture separates reads, writes and storage to reduce costs across all workload sizes and types. It uses vector clustering on top of blob storage for low-latency searches over data stores of nearly any size using purpose-built indexing and retrieval algorithms for fast and memory-efficient vector search. A multitenant compute layer supports thousands of users with on-demand availability.

Pinecone Serverless is now available in public preview on the Amazon Web Services Inc. cloud and will soon be released on Microsoft Corp.’s Azure and Google LLC’s Cloud Platform.

Pinecone claims to have over 5,000 customers and has raised $138M in venture funding.

Image: Pixabay