10 - Cloud Platforms
The managed AI stacks of the three hyperscalers and Databricks - model access, managed RAG, guardrails, agent runtimes, data platforms and MLOps - and how to translate experience on one into credible vocabulary on the others.
Learning objectives 3-4 hours
By the end of this module you will be able to:- Describe the model-access, RAG, guardrail and agent-runtime services of AWS Bedrock, Microsoft Foundry and Google's Gemini Enterprise Agent Platform
- Map any of these capabilities across clouds and name one architectural difference that matters
- Explain the Databricks lakehouse - medallion architecture, Delta Lake, Unity Catalog, MLflow - and its GenAI services
- Choose between a managed service and a self-managed path (GKE, GPU clouds, inference providers) for a workload
Prerequisites
What You Will Learn
- AWS Bedrock: Converse API for unified model access, Knowledge Bases for managed RAG, Guardrails, AgentCore, and cost/customization levers (batch, prompt caching, provisioned throughput, Custom Model Import)
- Databricks and Spark: the medallion architecture, Delta Lake, Unity Catalog, MLflow, GenAI services, and DBU cost levers
- Microsoft Foundry (formerly Azure AI Foundry): model catalog, Foundry Agent Service, evaluations, Microsoft Fabric vs BigQuery, Azure SQL/Cosmos DB
- Google Cloud's Gemini Enterprise Agent Platform (formerly Vertex AI): Model Garden, ADK, Agent Runtime, Memory Bank, governance
- How to translate experience on one cloud into credible vocabulary on the others
- A reusable Google Cloud ↔ AWS ↔ Azure comparison table, plus where GPU clouds and inference providers fit
Chapter Map
| # | File | Topic | Difficulty |
|---|---|---|---|
| 1 | AWS Bedrock | Converse API, Knowledge Bases, Guardrails, AgentCore | Intermediate |
| 2 | Databricks & Spark | Medallion architecture, Delta Lake, Unity Catalog, MLflow, PySpark, DBU pricing | Intermediate |
| 3 | Microsoft Foundry | Model catalog, Foundry Agent Service, Fabric vs BigQuery, Azure SQL, Cosmos DB | Intermediate |
| 4 | Google Cloud Agent Platform | Vertex AI's successor: Model Garden, ADK, Agent Runtime, Memory Bank, governance, GKE/TPU paths | Intermediate |
| 5 | Platform Comparison Table | Google Cloud ↔ AWS ↔ Azure mapping across models, agents, data and MLOps; GPU clouds | All levels |
| 6 | Q&A Review Bank | 19 Q&A pairs across all topics | All levels |
Recommended Learning Paths
Path A: Cloud-by-Cloud
- AWS Bedrock
- Databricks & Spark
- Microsoft Foundry
- Google Cloud Agent Platform
- Platform Comparison Table - tie it together
Path B: Interview Preparation (Accelerated)
- Platform Comparison Table - the fastest path to sounding fluent across clouds
- Q&A Review Bank - drill the specifics
- Whichever single-cloud note matches the role you're interviewing for
Path C: Databricks-Specific Roles
- Databricks & Spark
- Cross-reference Model Lifecycle & Rollout - MLflow is the concrete tool behind that note's registry concept
Resources
- Q&A Review Bank - 19 Q&A pairs in this module
- Module quiz - every Check Yourself question in this module
Key Cross-References
- Managed RAG concepts these platforms wrap → RAG Fundamentals
- Guardrail patterns Bedrock/Azure implement → Production Agents: Evaluation and Benchmarks
- Registry concept MLflow implements concretely → Model Lifecycle & Rollout
Section Appendix
Summary & Key Terms - a quick recap of this section and its essential vocabulary.
Next Topic
Previous: 09 - Production Engineering · Next: 11 - Prompt & Context Engineering