In this Specialization, you’ll learn how to plan, deploy, and optimize AI infrastructure on Google Cloud. You’ll explore how AI and high-performance computing workloads depend on the right deployment model, network design, storage architecture, and accelerator choice.
You’ll compare options for GPU-accelerated clusters, including Google Compute Engine and Google Kubernetes Engine. You’ll also examine how GKE supports inference workflows through containerization, networking configurations, distributed training, GPU sharing, and model-level optimization.
Across the networking and storage courses, you’ll connect each part of the AI pipeline from data ingestion to training, inference, serving, and archiving. You’ll explore Cross-Cloud Network, Cloud Interconnect, Jumbo Frames, RDMA, Titanium offload, GKE Inference Gateway, IAM, Cloud Storage, Anywhere Cache, Dataflux Dataset, Cloud Storage FUSE, Managed Lustre, Hyperdisk ML, GPUs, and TPUs.
By the end of this Specialization, you'll be able to:
Select deployment options for AI workloads using GCE, GKE, GPU clusters, and inference workflows.
Design networking and storage choices for AI data ingestion, training, serving, and archiving.
Compare GPUs and TPUs and apply optimization strategies for performance, efficiency, and flexibility.
Applied Learning Project
You’ll build applied AI infrastructure judgment through practical demonstrations, scenario-based lessons, and graded module-end quizzes. As you move through the program, you’ll learn how to create GPU-accelerated clusters, provision workloads on Google Compute Engine, configure Google Kubernetes Engine, and plan GKE-based inference deployments.
You’ll also assess networking choices for data ingestion, distributed training, and generative AI inference, including Cross-Cloud Network, Cloud Interconnect, Jumbo Frames, RDMA, GKE Inference Gateway, queue-depth routing, IAM, and reliability practices. Storage activities focus on matching services to each AI pipeline stage, including Cloud Storage, Anywhere Cache, Dataflux Dataset, Cloud Storage FUSE, Managed Lustre, Hyperdisk ML, checkpointing, model serving, and archiving. GPU and TPU modules help you compare accelerator options, provisioning choices, interoperability, and performance tuning strategies.
