quadrantthe company that develops the eponymous open-source vector database, has raised $28 million in a Series A round led by Spark Capital.
Founded in 2021, Berlin-based Qdrant seeks to capitalize on the burgeoning AI revolution. Open source vector search engine and database targeted at developers. It is an integral part of generative AI and requires drawing relationships between unstructured data (text, images, etc.). Even if the data is “dynamic” within a real-time application (such as unlabeled or otherwise unorganized audio).Street gartner dataunstructured data accounts for approximately 90% of all new enterprise data and is growing three times faster than structured data.
Vector database space is hot. Over the past few months, we’ve seen things like: Weaviate raises $50 million Meanwhile, Zilliz has secured $60 million to commercialize its open source vector database.Other chroma Secured $18 million in seed funding Meanwhile, Pinecone won $100 million for its own alternative.
Meanwhile, Qdrant raised $7.5 million last April, further highlighting the seemingly insatiable appetite investors have for vector databases, but also pointing to a planned growth spurt on Qdrant’s part.
“Our plan was to raise our next round of funding in the second quarter of this year, but when we received the offer a few months ago, we decided to save time and start scaling the company now,” said Qdrant CEO and CEO. Co-founder Andre Zayarny explained. Tech Crunch. “It always takes time to raise funding and hire the right people.”
Notably, Zayarni says that at the same time the company received a follow-on investment offer, it actually rejected a potential acquisition offer from a “major company in the database market.” “We made the investment,” he said, adding that the company will use the new cash injection to strengthen its business team, as the company is essentially made up of engineers at the moment.
binary logic
In the nine months since its last funding, Qdrant has launched A new ultra-efficient compression technology called binary quantization (BQ) focuses on low-latency, high-throughput indexing, which they say reduces memory consumption by up to 32x and improves search speed by about 40x.
“Binary quantization is a way to ‘compress’ a vector to its simplest representation with just 0s and 1s,” Zayarni says. “Comparing vectors becomes the simplest CPU instruction. It makes queries much faster and saves a lot of memory usage. The theoretical concept is not new, but the loss of precision We implemented it in very few ways.”
However, BQ does not work with all AI models, and it is entirely up to you to decide whether the compression option works best for your use case. However, Zayarni said his OpenAI model had the best results, and Cohere also worked. Google’s Gemini was similar. The company is currently benchmarking against models such as Mistral and Stability AI.
Such efforts have led to companies such as Deloitte, Accenture, and perhaps the most well-known of them all, X (maiden name twitter). Or, perhaps more accurately, his xAI from Elon Musk, who developed ChatGPT competitor Grok and debuted on the X platform last month.
Due to a non-disclosure agreement (NDA), Zayarni did not provide details on how X or xAI uses Qdrant, but it is reasonable to assume that they are using Qdrant to process real-time data. is. In fact, Grok uses a generative AI model called Grok-1, trained with data from the web and feedback from humans, and is (currently) closely aligned with Real-time data from posts can be incorporated into responses. This is what is known today as the Search Enhancement Generation (rug), Elon Musk has been publicly teasing such use cases over the past few months.
Qdrant does not reveal which customers are using the open source Qdrant entity and which are using its managed services, but many startups such as GitBook, VoiceFlow, and Dust are ” using that service. Managed Cloud Services – This allows companies with limited resources to benefit from not having to manage and deploy everything themselves, as is the case with core open source.
But even if companies choose to pay for add-on services, Zayarni is adamant that the company’s open source credentials are one of its key selling points.
“There is always a risk of vendor lock-in when using proprietary or cloud-only solutions,” Zayarni said. “If a vendor decides to adjust prices or change other terms, customers must agree to or consider moving to an alternative product, which is not easy if it is a high-volume production use case. . With open source, you always have more control over your data and can switch between different deployment options.”
In addition to today’s funding, Qdrant is also officially releasing a managed “on-premises” edition. This allows businesses to host everything in-house while still taking advantage of the premium features and support that Qdrant provides.This is: last week’s news The cloud edition of Qdrand has been introduced to Microsoft Azure, adding support for existing AWS and Google Cloud Platform.
In addition to lead backer Spark Capitali, Qdrant’s Series A round included participation from Unusual Ventures and 42cap.