Keera

The Challenge

Building Kubernetes platforms for clients has been bespinian's job for years. With Keera, the team launched a product of its own: an AI stack for Swiss companies that want to use AI assistants without their source code, prompts and customer data ending up with US corporations. That promise covers the whole stack, so running the gateway and web front ends on AWS or Azure was never an option. At the same time, the small team wanted to get Keera to market quickly, and setting up its own Kubernetes cluster would have taken time that belonged in the product. So for its own product, bespinian deliberately chose not to build the kind of platform it builds for everyone else.

The Solution

Keera runs on Deploio, our Swiss Platform-as-a-Service. The team pushes its code with Git, and Deploio builds, deploys and runs the application in our data centres in Switzerland, together with a PostgreSQL database and a Redis-compatible key-value store. The Keera Gateway connects to the models themselves through Swiss inference providers. Shortly after the start, Keera was live on its own domain and bespinian could begin testing. When questions come up, the team reaches us directly and gets technically sound answers without detours.

The Result

bespinian takes care of the product, not the platform: a release is a Git push. This means Keera's sovereignty promise also holds for how the gateway and front ends are run. By the team's own estimate, this saves considerable effort and cost compared with operating a Kubernetes cluster itself. bespinian handles most things through self-service, and for everything else we are a direct contact away.