Qdrant is developing a new type of database that works particularly well for artificial intelligence. With new millions, the startup is now set to become really big.
The creators of Qdrant do not describe themselves as AI experts. Qdrant
The database startup Qdrant has raised the equivalent of 26 million euros (28 million US dollars) in Series A. The money comes exclusively from the USA. The lead investor in the round is Spark Capital. The investors were convinced by the Berliners’ special database. This is used by AI companies, among other things, to get problems such as hallucinations under control or to provide long-term memory.
Actually, they didn’t even want to go into fundraising, says Qdrant founder André Zayarni in an interview with Gründerszene. But the investors’ interest was so great that they still took part. “In a few months we might have gotten money at a higher valuation,” Zayarni said. He added with a laugh: “I wanted to save myself the stress of going back into fundraising.”
Qdrant raised the seed round just a few months ago. The equivalent of seven million euros came into the Berliners’ account. In total, the database startup has raised around 33 million euros so far.
So far, the product has “almost grown on its own,” says Zayarni. The money will now be used primarily to hire employees in the areas of sales and marketing. The Qdrant team currently consists of 40 people, all of whom are developers. These are not only located at the headquarters in Berlin, but are spread around the world.
Read too
These five AI investors reveal how you can convince them with your pitch (and what you can’t do at all)
“We are not an AI company”
The new employees now also have the task of explaining the product. This is quite complex and not only has to do with artificial intelligence. “We are not an AI company,” says Zayarni. “We are an engineering company that offers an infrastructure product. We are building a database that can be used to build AI applications.”
In principle, Qdrant’s database can be used for many things. This is because it works with vectors, hence the name: vector database. To do this, the data must first be converted into vectors. These vectors can then be used to create relationships or similarities between objects. For example, a cat and a dog would be closer together than a cat and a person – or a car, according to the founder.
“This is the next stage of the search,” said the founder. This previously worked using certain words or keywords. A vector database, on the other hand, is about the meaning of words or entire sentences. You can use it to describe what you are looking for and you don’t necessarily have to use the right word.
Read too
“Working time almost halved” – This is how startups use ChatGPT in their daily business
Amazon relies on vector databases for search
The founder cites Amazon as an example, which has been using this type of search for years. If you search for coats there, you will also see products that do not contain the keyword at all, but similar products that come close to the meaning. This works not only with text, but also with audio or images. Similar image objects, music or voice are recognized.
Vector databases are not new, but Qdrant’s solution is particularly high-performance, according to the founder. Products that can be built with it are diverse – and range from facial recognition to product recommendations.
The fact that artificial intelligence works with a vector database is because neural networks – machine learning – also calculate with vectors. The AI hype has given the company a push, says the founder, so that artificial intelligence has been the most prominent use case since last year. The vector database should be able to solve the problem of hallucinations – fictitious things that chatbots present as facts – or allow access to private data. Chatbots like ChatGPT dock into the vector database and make queries.
Read too
Germany is making big mistakes when it comes to artificial intelligence – that has to change, says an expert
Qdrant in exchange with OpenAI, Mistral and Aleph Alpha
You get the AI expertise yourself through the partners. “We are in contact with everyone,” says Zayarni. He lists: We are in contact with OpenAI, Cohere, Mistral, Aleph Alpha and others. “Everyone is interested in the partnership because we are the link between the AI models and the users.”
Qdrant is open source and therefore free for the community participating there. For a year now, there has also been a paid cloud product that users do not have to install or configure themselves. Another product will soon be launched, a hybrid model in which large companies in particular have their databases running on their own servers. Qdrant itself is then responsible for administration as a cloud.
According to the company, the software was downloaded five million times last year. Customers include Deloitte, Hewlett Packard Enterprise and Bayer.