Vector database startup Pinecone Systems said today it raised $28 million in a series A round of funding to help build out its technology and go-to-market efforts.
The vendor, based in San Mateo, Calif., was founded in 2019 by Edo Liberty, who spent nearly seven years working at Yahoo on machine learning, followed by three years at AWS, where he worked on AI projects including Amazon SageMaker.
Liberty has a doctorate in computer science from Yale University, where his thesis was on random projection, a machine learning technique that has direct impact on vector databases.
With a vector database, rather than using just keywords to gauge relevance, content is converted into mathematical arrays. AI algorithms within the database help determine the nearness of vectors to assess query relevance.
In this Q&A, Liberty, who is CEO of Pinecone, talks about the vendor's vector database technology.
Why are you now raising a series A to build out the Pinecone vector database technology?
Edo Liberty: We raised a very large, healthy seed round of $10 million, slightly more than a year ago, and we had plenty of runway left. So we were not planning on raising at all. We launched our commercial product in October 2021, and the traction went through the roof.
The commercial product is a fully managed cloud service. Users query against data through their applications. They don't need to worry about machines and availability and backups and anything like that -- we take care of everything.
Edo LibertyFounder and CEO, Pinecone
The number of $28 million that we raised didn't come down from the mountain. It was based on our planning to help support the rapid growth that we have and the need for engineers, product managers, customer success and developer relations staff. We mapped it out for the next 18 to 24 months, and the number was roughly 28.
How is a vector database different from a graph database?
Liberty: Different databases specialize in different kinds of data. Graph databases predominantly care about relationships between objects. A social network is a standard example of a graph. In a social network, the relationships that you care about are pairwise relationships between objects in your data.
A vector database cares about vectors, which are just numeric arrays. It's the way that deep learning networks represent data. That data could be text, images or anything else.
The vector is a semantically rich representation of the objects, and that is what is saved in your data. So you query against that vector to work with the data. Deep learning models and machine learning models in general are numeric objects. If you want to apply math to something, you have to use numbers.
How can the Pinecone vector database work with traditional data structure for analytics and business intelligence?
Liberty: The capabilities of AI and vector search are inherently soft and fuzzy and hard to pin down for traditional data engineers doing BI and data analytics.
One of the most difficult things in building something like Pinecone is combining traditional data structure and traditional database capabilities with this advanced AI vector-based search, and we have done that.
Editor's note: This interview has been edited for clarity and conciseness.