Review a compact AWS Certified Generative AI Developer - Professional (AIP-C01) cheat sheet for Bedrock, RAG, agents, safety, governance, optimization, evaluation, and production troubleshooting before using IT Mastery practice.
Use this cheat sheet to keep professional-level GenAI decisions organized before you take the free AIP-C01 diagnostic. The exam rewards production judgment: grounding, data boundaries, safety, integration, evaluation, and operational trade-offs.
| Item | Review cue |
|---|---|
| Exam route | AWS Certified Generative AI Developer - Professional |
| Exam code | AIP-C01 |
| Items | 75 total, including scored and unscored items |
| Time | 180 minutes |
| Practice option | Live IT Mastery practice available |
| Best use | Practice production GenAI architecture, implementation, security, operations, and troubleshooting decisions |
| Domain | Weight | What to know | Common trap |
|---|---|---|---|
| Foundation model integration, data, and compliance | 31% | Bedrock model access, RAG, knowledge bases, data handling, compliance boundaries | sending sensitive data through a path with weak controls |
| Implementation and integration | 26% | agents, tool use, APIs, orchestration, streaming, application integration | choosing a complex agent when a simpler prompt or retrieval flow fits |
| AI safety, security, and governance | 20% | guardrails, IAM, encryption, logging, human review, output controls | treating model output risk as only a prompt problem |
| Operational efficiency and optimization | 12% | latency, token cost, throughput, caching, model choice, batching | optimizing cost by weakening quality or safety |
| Testing, validation, and troubleshooting | 11% | evaluation sets, retrieval diagnostics, prompt defects, permissions, monitoring | blaming the model when retrieval or permissions failed |
| Distinction | Exam reflex |
|---|---|
| Prompt issue vs retrieval issue | Bad instructions point to prompt design. Missing or wrong source context points to retrieval. |
| RAG vs fine-tuning | Use RAG for current or private facts. Use fine-tuning only when model behavior itself must adapt. |
| Guardrail vs IAM | Guardrails shape outputs. IAM controls access to AWS resources. |
| Agent vs workflow | Agents decide tool use. Workflows are more deterministic and easier to govern. |
| Evaluation vs monitoring | Evaluation tests expected behavior. Monitoring watches production behavior and drift. |
| Token cost vs latency | Smaller prompts and models can help both, but quality and safety still control the final decision. |
When you miss an AIP-C01 question, write the failed layer: grounding, implementation, safety, efficiency, or troubleshooting. Then drill that domain before another full mixed run. Professional-level GenAI readiness is the ability to explain the trade-off, not just identify a Bedrock feature name.