Pinecone to Bolster Vector Search Engineering Team Following Series A
Pinecone Systems, a pioneer in enabling vector search, will be bolstering its engineering team to solidify its early momentum in the booming vector database business following its $28 million Series A, the company told Datanami.
“Building great databases is hard, and if you want to build the best database out there, you have to keep investing very deeply in the technology and the product,” Pinecone Systems CEO Edo Liberty said in an interview last week. “We will definitely keep doing that.”
The company is enjoying significant success with the free tier of its vector database, which can be used to power a vector search engine supporting millions of items. The company is converting a significant number of those free users to the paid version, which enables AI techniques to be used to power search against billions of objects.
Pinecone is growing at a 20% monthly clip in terms of revenues and customers, Liberty said. With such lofty revenue growth, there’s less of a need to invest in sales and marketing. Instead, more than half of the funding of the $28 million Series A, which it announced today, will go towards keeping the company’s position in the booming vector search market, particularly as big players begin to enter it, the CEO said.
“We’ll also grow the marketing arm and developer relations, customer success, sales, and so on for sure,” Liberty said. “But [there will be] a huge focus on helping customers get better results–better search, faster, easier, cheaper. Definitely a huge investment.”
One might be tempted to think of Pinecone as an overnight success in the suddenly hot vector search market, which is sometimes called neural search because of the way neural networks are used to power the search engine instead of traditional keywords. But Liberty says it’s the culmination of a lot of hard work to fulfill his team’s insight.
“That wave is hot now, but we started three years ago with what was then a vision,” he said. “And to be honest, most people that I spoke with either thought I’m nuts or stupid. They had no idea what the hell I was talking about. So I’m glad that the market and reality has validated our vision, and now it’s hot and trending and now it’s obvious, but it wasn’t obvious all along.”
Liberty and his distributed team developed a vector database from scratch using the language Rust, which he said gives Pinecone an advantage over other companies that are taking a quicker but less comprehensive path into the world of vector search and vector databases.
“Any company who goes into this is going to have a hell of a time building something that’s actually worth using,” he said. “That said, I have no doubt that the big players, if they come in, they’re going get some traction. I hope it’s going to be like a high tide raising all boats kind of situation. The market is so massive and the penetration of this entire idea and entire product category is still nascent.”
Pinecone, which has offices in New York City and Tel Aviv, Israel, is seeing competition emerge on several fronts. There are existing search engine companies beginning to augment their offerings with vector search capabilities. For example, Elastic recently bolstered its Elasticsearch offering with vector search capabilities.
Pinecone is also seeing vector indexes being offered as open source libraries that users can plug into their existing systems. These systems enable users to benefit from approximate nearest neighbor search capabilities, Pinecone Vice President of Marketing Greg Kogan said, but they don’t offer the full benefits of a database.
Lastly, application vendors and service providers are beginning to offer vector search capabilities along with traditional keyword search that are built-in to existing offerings, such as with cloud giants. This will be popular with organizations that typically buy out-of-the-box capabilities, Kogan said.
“We’re not saying that to be an effective database you have to check these six boxes and everyone else falls short,” Kogan said. “They’re all providing vector search capabilities. It’s just to what extent. The features, scalability, the performance, and cost make a difference, even though they can all say they all do better searching.”
Pinecone is targeting bigger companies that have larger search workloads that can benefit from having a full database engine underneath the hood that provides full CRUD support, as well as other database capabilities. The company’s cloud-based model provides those capabilities via a simple API, which helps to keep the technical muss and fuss to a minimum.
“We follow our customers very, very closely. They tell us exactly what they need,” Liberty said. “They need the search to be better, so they want it to be vector-based to begin with. They want filters and all sorts of lower-level control over the results they get. They want it to scale. They want it to be cost efficient. And they wanted to be rock solid and hands free.”
In Liberty’s view, AI search will eventually win out over the more rudimentary keyword approach. Except for a few areas, such as what can be powered with a super-fast key-value store, the better context provided by vector search will win out, he said.
“We just want to focus on our customers and say, hey, when we are there right tool that you should use us, when something else is the right tool, use something else,” he said. “We are very, very good at what we do.”
Editor’s note: This article has been corrected. Pinecone Vice President of Marketing Greg Kogan’s name was misspelled. Datanami regrets the error.