Blog

Notes from the team.

What we are learning while building sovereign platforms, written by the engineers doing the work, not a content team.

Gemini Generated Image 152pr0152pr0152p

Latest

Kubernetes offers namespaces or dedicated clusters per tenant. We look at the isolation patterns that sit between those two extremes.

·

Dominik Heilbock

DSC05275

·

Max Körbächer

Four days at the SAP Innovation Center in Potsdam, hacking on sovereign inference and AI engineering with the ApeiroRA and NeoNephos projects.

Bildschirmfoto 2026 07 29 um 14 26 02

·

Jakob N.

An internship project: building a local AI chatbot that runs entirely on your own machine, and learning what really happens behind ChatGPT.

Thumbnail streaming tokens in bottleneck

·

Lino Deppe

Comparing vLLM deployment strategies for production: standalone, KServe and BentoML, with canary rollouts and where each fits in LLMOps.

·

Mohit Gupta

Namespace-per-team is the default way to share a GPU cluster. It's also the wrong one. This is what true multi-tenancy looks like in Terraform.