Practice lab · 2026 · Practice

Cloud-native lab

Problem

AI services need to be packaged, deployed and scaled reliably.

Solution

A FastAPI service in a multi-stage Docker image, deployed to a local Kubernetes cluster with role-based access, Helm and autoscaling, including scale-to-zero.

Stack

  • Docker
  • Kubernetes
  • Helm
  • KEDA
  • FastAPI
  • uv

Engineering

  • Multi-stage uv build
  • RBAC roles tested with auth can-i
  • HPA and KEDA autoscaling