$ run as a normal user# run as root (or with sudo)
Know why MLOps comes next
Companies train and run more AI models every year. Someone must build the platform: GPUs on Kubernetes, model pipelines, model registries, serving and monitoring. That is MLOps — DevOps for models.
Study with these sources
All main courses are free.
| Source | Use it for |
|---|---|
| Google Machine Learning Crash Course | ML basics before MLOps |
| ★ MLOps Zoomcamp — DataTalksClub | A complete free course: tracking, orchestration, deployment, monitoring |
| ★ Made With ML — MLOps course | Production ML from design to serving |
| LFS147 Kubeflow intro (free) | AI and ML on Kubernetes |
| Designing Machine Learning Systems — Chip Huyen | The book |
| My summary | MLOps |
Build a first MLOps project
One small model, the full path: experiment tracking with MLflow, a training pipeline, a container with the model behind an API, deployed on the home cluster by GitOps, with a dashboard for the model's answers.
Know the later specialisations
After 2 – 3 years as Cloud / DevOps Engineer, choose a direction.
| Direction | Focus |
|---|---|
| Cloud Security Engineer | Identity, policies, threat detection, compliance in the cloud |
| Platform Engineer / Architect | Internal developer platforms, Kubernetes at scale, golden paths |
| MLOps / AI Infrastructure | GPU clusters, model pipelines, serving and monitoring |