AI/ML Storage Design

This skill helps design storage architecture for AI and ML workloads on Google Cloud, covering both Training and Serving (Inference) phases.

0. Critical First Step: Check Existing Infrastructure

Before proposing new infrastructure, ALWAYS ask the user if they have existing storage resources. This prevents redundancy and leveraged existing investments.

Ask:

• "Do you already have a Managed Lustre instance running for your training pipeline?"
• "Are your datasets or models already stored in Cloud Storage buckets?"

List Lustre instances:

gcloud lustre instances list --location=<REGION>