Qdrant, the developer of a high-performance, open-source vector database, today announced it has raised $28 million in early-stage funding led by Spark Capital enhance its offering for generative artificial intelligence use cases.

Existing investors, Unusual Ventures and 42CAP also participated in the company’s Series A funding round. The fundraise follows Qdrant’s $9.8 million seed funding round in April 2023, which brings the total funding for the company to $37.8 million.

Generative artificial intelligence models, such as the highly popular ChatGPT chatbot from OpenAI, require complex unstructured data such as text, images, audio and video to power them in order to provide conversational question and answer capabilities, search and summarization. Simple keyword searches with unstructured data don’t work as well when the contextual relationships within the information are just as important as the data being searched.

“Vector databases are designed to handle complex high-dimensional data, unlocking the foundation for pivotal AI applications,” said Andre Zayarni, co-founder and chief executive officer of Qdrant. “They represent a new frontier in data management, in which complexity is not a barrier but an opportunity for innovation.”

Vector databases allow large language models, the AI models that chatbots are built on, to remember contextual relationships over time, thereby giving them a memory. They are also fundamental to improving the reliability of generative AI models by storing large amounts of real-time authoritative contextual data for retrieval-augmented generation, or RAG.

Both of these techniques are fundamental for reducing hallucinations, where an AI confidently makes a completely false statement. RAG is particularly useful for enterprise applications where AI models might have trouble answering expert-level questions about industry-, company- or product-specific information. In these situations, reducing the chances that an AI answers a question incorrectly when it comes to industry or company relevant information can be critical.

Zayarni said that Qdrant sets itself apart from competitors by creating a highly performant and scalable vector database product on Rust, the systems programming language, for memory safety and scale. The company also deployed a custom filtering algorithm that allows for rapid vector search with very low latency across billions of vectors.

Last week, the company expanded Qdrant Cloud, its managed vector database offering onto Microsoft Inc.’s Azure cloud. This will allow customers to rapidly set up their own environments in Azure and reduce deployment time. That adds to the company’s already existing Amazon Web Services and Google Cloud Platform support.

In addition to the funding, Zayarni announced the launch of an on-premises enterprise edition that will allow customers to host their own vector databases in-house for higher data security that still have the same features available as the ones with managed support. For customers looking for further flexibility, Qdrant also now offers hybrid software-as-a-service solutions for vector database support.

Image: Pixabay