Weaviate builds a vector database that teaches machines to understand what data means

Imagine searching a large document library for everything related to 'reducing energy costs'. A traditional search engine returns only results that literally contain those words. Weaviate also returns passages about insulation, heat pumps and smart meters, because the system understands what you mean, not just what you type.

That, in a nutshell, is what Amsterdam-based Weaviate does. The company, founded in 2019 by Bob van Luijt and Etienne Dilocker, builds an open-source vector database that stores and searches data on the basis of semantic meaning. Not keywords, but context. Not exact matches, but understanding.

With a Series B round of over €45 million, closed in early 2024, the company has the resources to develop that ambition further. But the core of Weaviate has remained the same for years: a database that supports AI applications at a scale reaching billions of objects.

From keyword to meaning

Traditional databases store data as rows and columns and search for exact or partial text matches. Vector databases work differently. Text, images or other data are converted into mathematical representations, so-called vectors, that capture the meaning of a piece of information across a space of hundreds or thousands of dimensions. Two sentences that mean the same thing but use different words end up close to each other in that space.

Weaviate is built around that principle. Developers can store data objects in it, equipped with vector representations, and then search them by semantic proximity. The platform handles vectorisation, result ranking and automatic infrastructure scaling itself. Anyone building an AI application therefore does not need to build or manage that technical layer themselves.

The founders and their background

Bob van Luijt, CEO of Weaviate, has a background in software development and machine learning. He combines technical depth with a broad interest in open-source and digital product development. Co-founder and CTO Etienne Dilocker comes from the world of distributed systems and cloud technology, a background that is directly visible in how Weaviate is structured: scalable, distributed and designed for production environments under heavy load.

Together they founded SeMI Technologies, the company that later put Weaviate forward as its product and brand name. The choice for open-source was deliberate from the outset. An open database attracts a community of developers who contribute to the code, report bugs and deploy the product in all manner of unexpected contexts. That accelerates development and builds trust among organisations that want to avoid proprietary vendor lock-in.

What you can build with it

Weaviate positions itself as infrastructure for three types of AI applications that are currently gaining significant traction. The first is semantic search, enabling users to search through large volumes of documents, images or other content in natural language. The second is Retrieval Augmented Generation, commonly abbreviated as RAG. In RAG, a language model is connected to an external knowledge source: the database supplies relevant context, after which the model generates a response based on that specific information. This reduces the risk of fabricated answers and makes it possible to use language models over company-specific data.

The third category is that of agentic workflows, in which AI systems execute multiple sequential steps, retrieving information, reasoning and taking actions. For this too, a fast, reliable vector database is an essential component of the architecture.

Weaviate supports integrations with common AI frameworks and language models, allowing developers to connect the platform to the tools they already use.

Open-source as strategy

The open-source core of Weaviate is not a side note. It is a deliberate choice that enables the company to be widely adopted, from individual developers experimenting locally to large enterprises running the database in their own cloud environment. Alongside the self-hosted version, Weaviate offers a managed cloud variant, a common revenue model in the open-source database world.

This approach has earned Weaviate an active user community and contributed to the project's visibility in the broader AI infrastructure market. It is a market that has grown rapidly in recent years, driven by the rise of large language models and the need for a specialised storage layer for vector data.

Amsterdam as a base for AI infrastructure

Weaviate is one of the few European players in a market that is currently dominated by American companies such as Pinecone and Chroma. Operating from Amsterdam, the company competes globally with a product that aligns with the needs of organisations seeking to retain control over their data and infrastructure, both technically and in terms of its licensing model.

For the Dutch and European AI scene, Weaviate demonstrates that deep technical infrastructure can indeed be built and funded here. Against the backdrop of growing emphasis on digital sovereignty and European AI regulation, there is room for European alternatives in the infrastructure layer beneath AI applications. The fact that Weaviate operates from Amsterdam while scaling globally shows that this combination is achievable.