1. First Principles of Generative AI on AWS

Before you can architect production GenAI systems on AWS, you need accurate mental models of how these systems actually work. Candidates who rush past first principles spend Phase 2 onwards memorizing services without understanding why those services exist — and they fail scenario questions that require reasoning from principles rather than pattern matching.

The professional-level exam is explicitly testing your ability to handle compounding requirements: a question might demand a solution that's simultaneously low-latency, cost-efficient, data-residency compliant, and hallucination-resistant. You can't solve that by recalling service names. You solve it by understanding the underlying constraints that each service addresses.

⚠️ Common Misconception: Generative AI on AWS is primarily about knowing which Bedrock model to call. In reality, the exam tests the system design around the model — data pipelines, retrieval architecture, safety controls, agent orchestration, and monitoring. The model itself is often the smallest decision.