Highlights:

  • Pinecone Systems Inc. highlighted that the cost of storing and searching through substantial volumes of vector data on-demand can be excessively high, even when utilizing a vector database.
  • Pinecone employs vector clustering on top of blob storage, enabling low-latency searches over data stores of nearly any size.

Pinecone Systems Inc., a well-funded startup specializing in vector databases, has unveiled a serverless iteration of its product explicitly designed for artificial intelligence applications.

According to the company, the serverless architecture — characterized by a cloud computing execution model in which the cloud provider dynamically manages the allocation and provisioning of servers — has the potential to slash costs by as much as 98%. Moreover, it prevents developers from needing to handle infrastructure provisioning, facilitating a quicker time-to-market for applications.

Pinecone Systems Inc. pointed to its research, revealing that expanding the data accessible for context retrieval results in a 50% decrease in unhelpful answers—responses that either inaccurately address or fail to address a given question. The company emphasized that developers of large language models can enhance the quality of generative AI models by simply increasing the accessibility of more data.

In contrast to relational databases, which organize data in rows and columns, vector databases encode unstructured data as high-dimensional data points. Each of these points represents a vector or an array of numbers. A key function of a vector database is to conduct similarity searches, enabling rapid identification of vectors that closely resemble a specified query vector. This process employs metrics such as cosine similarity or Euclidean distance to determine similarity.

The company highlighted that the cost of storing and searching through substantial volumes of vector data on-demand can be excessively high, even when utilizing a vector database. Pinecone’s technology is designed to store and search through AI-generated representations of unstructured data. These representations encapsulate the meaning of the original content in a machine-readable format, significantly enhancing efficiency for keyword-based searches.

The serverless architecture streamlines operations by eliminating the infrastructure provisioning stage, saving significant time and cost.

Furthermore, Pinecone’s architecture segregates reads, writes, and storage functions, reducing costs across various workload sizes and types. Pinecone employs vector clustering on top of blob storage, enabling low-latency searches over data stores of nearly any size. This is achieved through purpose-built indexing and retrieval algorithms, ensuring fast and memory-efficient vector search capabilities. The multitenant compute layer supports thousands of users, ensuring on-demand availability as needed.

Pinecone Serverless is currently accessible in public preview on the Amazon Web Services Inc. cloud, with forthcoming releases planned for Microsoft Corp.’s Azure and Google LLC’s Cloud Platform.

Pinecone asserts a customer base exceeding 5,000 and has secured USD 138 million in venture funding.