Exam at a glance

What you're taking

Exam code
AIP-C01
Full name
AWS Certified Generative AI Developer — Professional
Passing score
750 / 1000 (compensatory)
Questions
85 total · 75 scored · 10 unscored
Time
170 minutes
Question types
Multiple choice · Multiple response · Ordering · Matching
Cost
$300 USD
Validity
3 years

Domain weights

5 DOMAINS
D1 · Foundation Models 31%
D2 · Integration 26%
D3 · Safety & Security 20%
D4 · Optimization 12%
D5 · Testing 11%
Exam strategy Domains 1 + 2 = 57% of scored content. Master these first. Domain 3 plays to your CISSP/CCSP strengths — lean into that confidence. D4 and D5 are smaller but compensatory scoring means weak domains still hurt your total.

Domains

Each domain page walks through every task and skill from the official exam guide, with exam angles, AWS service mappings, and CSS architecture diagrams for the hardest concepts.

31%
Domain 1

Foundation Model Integration, Data Management & Compliance

Model selection, vector stores, RAG, embeddings, chunking, prompt engineering & governance. The biggest domain.

26%
Domain 2

Implementation and Integration

Agentic AI, model deployment, enterprise integration, FM API patterns, streaming, CI/CD for GenAI.

20%
Domain 3

AI Safety, Security & Governance

Guardrails, defense-in-depth, PII protection, prompt injection mitigation, responsible AI, compliance frameworks.

12%
Domain 4

Operational Efficiency & Optimization

Token economics, caching layers, model cascading, throughput optimization, GenAI observability.

11%
Domain 5

Testing, Validation & Troubleshooting

LLM-as-a-Judge, RAG evaluation, model evaluations, hallucination detection, regression testing.

10
Reference

Architecture Patterns

The 10 high-value architecture patterns the exam tests — each one diagrammed, with when-to-use and AWS service mappings.

Study tools

📋 Service Cheat Sheet

Every in-scope AWS service with its primary GenAI use case. Printable one-pager for the final-day review.

📖 Key Terms Glossary

Every acronym, concept, and technique the exam can throw at you, defined crisply.

⚠️ Question Patterns & Traps

How AIP-C01 questions are written, what distractors look like, and the traps that catch people who know the material.

🎯 Practice Questions

Six interactive question sets covering all domains, plus ordering & matching drills. Score tracked locally.

🧠 Recall Flashcards

Active recall on the highest-leverage facts. Mark known / unknown; focus repeats on weak cards.

📅 Final-Week Playbook

Day-by-day study cadence for the last week, exam-day morning routine, and pacing strategy.

Services to know cold

Not exhaustive, but these appear again and again. Full coverage on the cheat sheet.

Foundation models & RAG

  • Amazon Bedrock (core)
  • Bedrock Knowledge Bases
  • Bedrock Prompt Management
  • Bedrock Prompt Flows
  • Bedrock Cross-Region Inference
  • Amazon Titan (embeddings, FMs)
  • SageMaker AI (Registry, Endpoints, JumpStart)

Agents & integration

  • Bedrock Agents + AgentCore
  • Strands Agents (open-source)
  • AWS Agent Squad
  • MCP (Model Context Protocol)
  • Step Functions, Lambda
  • API Gateway, EventBridge, SQS
  • AppConfig (feature flags / model routing)

Vector & data

  • OpenSearch (Neural plugin, k-NN)
  • Aurora + pgvector
  • DynamoDB (metadata + vectors)
  • Kendra, S3, Glue (Data Quality, Catalog)
  • Comprehend (PII, entities, intent)
  • Textract, Transcribe, Rekognition

Safety & security

  • Bedrock Guardrails
  • Macie (PII discovery)
  • IAM, KMS, Cognito
  • VPC Endpoints, PrivateLink
  • CloudTrail, Bedrock Invocation Logs
  • WAF, Secrets Manager

Ops, cost, observability

  • CloudWatch (Logs Insights, metrics)
  • X-Ray
  • Cost Explorer, Cost Anomaly Detection
  • Managed Grafana
  • Bedrock Model Invocation Logs

Testing & quality

  • Bedrock Model Evaluations
  • SageMaker Clarify (bias, explainability)
  • SageMaker Model Monitor
  • SageMaker Ground Truth (labeling)
  • Amazon Augmented AI (A2I — human review)
Not on the exam Model development & training from scratch · advanced ML techniques · data engineering and feature engineering. This exam is about integrating and deploying FMs, not building them. If a question sounds like it wants you to train a model, look for the answer that uses a pre-trained FM or fine-tunes with LoRA/adapters instead.