Google's official Vertex AI cookbook: 735 stars, zero fluff

What it does

This is Google’s canonical sample repository for Vertex AI — notebooks, code snippets, sample apps, and structured “skills” for building and deploying ML and generative AI workflows on GCP. The README is upfront about its purpose: getting you from zero to a running model, whether via AutoML, custom training, or pulling from Model Garden.

The interesting bit

The “skills” directory is the unusual part — it’s not just notebooks, but a routed command reference for AI agent-style interactions with Vertex AI, covering the Gen AI SDK across five languages, deployment, inference, and fine-tuning. That’s a different shape than the typical “dump of Jupyter files” approach.

Key highlights

• Official notebooks plus a community/ directory for external contributions
• One-click run in Colab, Colab Enterprise, or Vertex AI Workbench
• Model Garden samples covering Gemini, Gemma, Llama 3, Claude 3, and others
• Structured “skills” with language-specific guides (Python, JS/TS, Go, Java, C#)
• Covers the full lifecycle: Feature Store, AutoML, custom training, Ray clusters, pipelines, prediction, explainability, metadata tracking

Caveats

• The repository explicitly disclaims being an officially supported Google product
• Generative AI notebook samples are actively being moved to a separate repo, so this isn’t the single source of truth for that slice

Verdict

Worth bookmarking if you’re committed to Vertex AI specifically and want working code over documentation. Skip it if you’re platform-agnostic or looking for deep conceptual explanations — this is samples, not strategy.