This document provides an overview of architecture guides to design, build, anddeploy AI and ML applications.

To help you find the right guidance that's relevant to your persona and needs, we provide the following types of architecture guides:

• Design guides: Prescriptive, cross-product guidance to help you plan and design your cloud architecture.
• Reference architectures: Detailed architecture examples and design recommendations for specific workloads.
• Use cases: High-level architecture examples to solve specific business problems.
• Deployment guides and Jump Start Solutions: Step-by-step instructions or code to deploy a specific architecture.

Agentic AI

Agentic AI applications solve open-ended problems through autonomous planning and multi-step workflows.

To build agentic AI applications on Google Cloud, start with the following guides:

• Design guide:Choose your agentic AI architecture components
• Design guide:Choose a design pattern for your agentic AI system
• Reference architecture:Multi-agent AI system in Google Cloud
• Explore moreagentic AI architecture guides.

Generative AI

Generative AI applications let use AI to create summaries, uncover complex hidden correlations, or generate new content.

To build generative AI applications on Google Cloud, start with the following guides:

• Design guide:Deploy and operate generative AI applications
• Design guide:Choose models and infrastructure for your generative AI application
• Reference architectures:Generative AI with RAG
• Explore moregenerative AI architecture guides.

ML applications and operations

Robust machine learning operations (MLOps) is the foundation for every AIinitiative, from classification and regression models to complex generative AIand agentic AI systems.

To build and operate ML applications on Google Cloud, start with the following guides:

• Design guide:Best practices for implementing machine learning on Google Cloud
• Blueprint:Build and deploy generative AI and machine learning models in an enterprise
• Reference architecture:Build an ML vision analytics solution with Dataflow and Cloud Vision API
• Reference architecture:Cross-silo and cross-device federated learning on Google Cloud
• Explore moreML applications and operations architecture guides.

AI and ML infrastructure

The performance, cost, and scalability of your AI and ML applications dependdirectly on the underlying infrastructure. Each stage of the ML lifecycle hasunique requirements for compute, storage, and networking.

The following resources help you design and select an appropriate infrastructure foryour AI and ML workloads:

• Design guide:Design storage for AI and ML workloads in Google Cloud
• Reference architecture:Optimize AI and ML workloads with Cloud Storage FUSE
• Reference architecture:Optimize AI and ML workloads with Google Cloud Managed Lustre