# 08 - AI and GPU Workloads on Kubernetes: Device Plugin and Training Jobs

In 2026, 66% of AI inference clusters run on Kubernetes. Learn how to configure the NVIDIA Device Plugin, schedule GPU training jobs, use Karpenter for spot scaling and optimize GPU costs by 40%.

## What you'll learn

- Installazione e configurazione del NVIDIA Device Plugin per Kubernetes
- Schedulare Pod con richiesta GPU (nvidia.com/gpu resource)
- Training distribuito con PyTorchJob e TFJob (Kubeflow Training Operator)
- Karpenter NodePool per provisioning automatico di nodi GPU spot
- GPU time-slicing per condividere GPU tra piu Pod

*This article is part of the **Kubernetes Avanzato** series on federicocalo.dev.*

---

## Read the full article

The complete article (20 min read) with code examples, diagrams, and practical exercises is available here:

**➡️ [08 - AI and GPU Workloads on Kubernetes: Device Plugin and Training Jobs](https://federicocalo.dev/en/blog/kubernetes-ai-gpu-workload-device-plugin-training)**

`https://federicocalo.dev/en/blog/kubernetes-ai-gpu-workload-device-plugin-training`

---

*By [Federico Calò](https://federicocalo.dev) — Software Developer & Technical Writer*