Inovai / About us
We’re preparing our own GPU server in Czechia. Your data stays out of the cloud.
We’ll run local models, fine-tuning and attack testing on our own hardware. Sensitive data won’t have to go to cloud APIs.
Local modelsfor clients, without cloud APIs
Fine-tuningan open model for your task
Attack testingon open models, in Czech and English
Sensitive documentsextraction and OSINT analyses
Data centre in Czechia, without cloud APIs
Problem
Sensitive data must not go to a cloud API and you don’t have your own GPUs yet.
What we’ll do
We run, fine-tune and test the model on our server in Czechia.
Result
You know where the data is. You can move the model to your premises later.
What we’ll use it for
Four things that are hard or expensive without your own hardware.
Local models for clientsThe model will run on our server and data won’t go to cloud APIs. Suitable for a pilot and for production.
Fine-tuning for your taskWe fine-tune an open model on your data. On a narrow task this closes some of the gap to large cloud models; we measure how much.
Attack testing and the Czech indexWe’ll attack open models locally, without the limits of cloud services. The measurements for the Czech index will run on it too.Sensitive documents and OSINTContracts, internal documents and OSINT analyses you don’t want to send to the cloud.Where will the server be?
In a data centre in Czechia. Data won’t leave the Czech Republic.
Is your own server better than the cloud?
Not always. The largest cloud models are stronger at general reasoning. Your own server makes sense when data must not leave, when you want to fine-tune a model or when we attack one.
How large a model can it run?
Open models up to roughly 70 billion parameters. What your task actually needs, we measure on your data.
Can the model eventually run on our premises?
Yes. You can start on our server and measure the results. We then move the model to your hardware and test it again.