How googlecloudplatform/applied-ai-engineering-samples works
Project type and purpose
• This is a monorepo of reference blueprints, code samples, and hands-on labs for “Applied AI Engineering” on Google Cloud.
• It is divided into three main verticals:
• Generative AI on Vertex AI (“genai-on-vertex-ai”)
• AI/ML infrastructure best practices (“ai-infrastructure”)
• Research operationalization examples (“research-operationalization”)
File structure
• Top-level directories:
• .github, docs, assets – repository management & documentation.
• ai-infrastructure – Terraform modules and end-to-end environment definitions for provisioning GCP infra (GKE, TPU, Cloud Storage, etc.).
• genai-on-vertex-ai – notebooks, scripts, and sample apps demonstrating Vertex AI services: foundation models (Gemini), evaluation, RAG, search, agents, prompting recipes, “needle in a haystack” tool.
• research-operationalization – code and notebooks showing how to run research models (e.g., timesfm) on Vertex AI.
• Under ai-infrastructure:
• terraform-modules – reusable modules (bootstrap, gke-aiml, kueue-config, metrics-tracking, workload-identity).
• tpu-training-on-gke – step-by-step environment folder (0-bootstrap → 1-base-infrastructure → 2-gke-config) plus examples.
• Under genai-on-vertex-ai:
• gemini, vertex_evaluation_services, retrieval_augmented_generation, vertex_ai_search, agents, research prompts, each with its own notebooks, scripts, configs.
• Under research-operationalization:
• timesfm folder with a Jupyter notebook.
README for additional insights
• The root README declares the three pillars and links to each section’s README for deeper architecture descriptions.
• The ai-infrastructure README will describe the cloud architecture (Terraform → GCP resources).
• The genai README sections describe which Vertex AI services are exercised.
Detail
• Essential element: local developer, CI, Terraform, GCP infra, code workloads, Vertex AI services, storage.