Our mission
HostYourAI exists to give European organisations sovereign AI infrastructure: a way to run open language models without depending on US cloud platforms. Data about patients, citizens and customers belongs under European law, on servers that physically sit in Europe, operated by a company you can actually reach.
Our story
HostYourAI was founded by Jantine Doornbos in Groningen, the Netherlands. It started as a practical problem in her own engineering work: the strongest language models were only available through US APIs, while the projects that needed them most handled data that was not allowed to leave Europe. Open models like Llama and Mistral already existed, but sourcing GPUs, configuring vLLM and keeping everything compliant turned out to be the same heavy lift for every single team. So we standardised that lift: first for our own projects, then as a platform for others.
EU Inference Router
‹ Insight 1 of 28 ›Why Groningen
We work from Groningen on purpose. The city has a strong technical university, a growing IT scene and a down-to-earth culture that prefers finished software over big promises. For a company whose product is reliability and trust, that is a good home base. The team is small: a handful of engineers who build, operate and support the infrastructure themselves. When you email us, the reply comes from someone who also ships code to the system you are running on.
What the platform is today
HostYourAI consists of four parts that work as one. The EU Inference Router is a shared, OpenAI-compatible API where you pay per token and can start right away. Dedicated GPU instances serve teams that need isolation or predictable capacity, billed per hour. The Model Garden lists more than a hundred open models, from Llama and Qwen to Mistral and DeepSeek, each with its current availability. And the Playground lets you test any model without writing a line of code.
Everything runs on GPUs in European datacenters, served with vLLM. We do not train on customer data, a Data Processing Agreement is available and our subprocessor list is public. That is not a marketing claim, it is simply how the system is built.
How we work
We would rather grow slowly and stay dependable than move fast and wobble. Features go live only when we trust them in production ourselves, and our pages describe what the platform can genuinely do today. Questions, ideas, or just curious who is behind this? Email info@hostyourai.com and a real person will answer.
