Gke Inference

If you're serving AI models on GKE, this walks you through Google's Inference Quickstart tooling to generate optimized Kubernetes manifests for LLM deployment. It covers the discovery workflow for matching models like Gemma or Llama with the right GPU or TPU hardware, then shows how to deploy with vLLM, TGI, or other inference servers. The accelerator selection table is genuinely helpful, from budget L4 GPUs up to the new Blackwell chips. Includes practical autoscaling configs using DCGM GPU metrics and optimization tips around quantization and KV cache tuning. Most useful if you already have a GKE Autopilot cluster and need to go from zero to deployed inference endpoint without reinventing the wheel.

Install to Claude Code

npx -y skills add google/skills --skill gke-inference --agent claude-code