CLF-C02 exam: what you need to know about generative AI
· 2 min read
AWS has increased the weight of generative AI topics in the CLF-C02 exam. Here's what's testable now.
What's in scope
The exam now explicitly tests:
- Amazon Bedrock — fully managed service to build generative AI apps using foundation models (FMs) from AI21 Labs, Anthropic, Cohere, Meta, Mistral, and Amazon
- Amazon Q — generative AI assistant for business (Q Business) and for developers (Q Developer, formerly CodeWhisperer)
- Amazon SageMaker — for building, training and deploying ML models; exam focuses on the managed aspect, not the data-science workflow
- AI/ML managed services — Rekognition (images/video), Comprehend (NLP), Translate, Polly, Transcribe, Lex, Forecast, Personalize
Key distinctions the exam likes to test
| Service | What it does | Common distractor |
|---|---|---|
| Bedrock | Access third-party FMs via API, no ML expertise needed | SageMaker (requires ML expertise) |
| Amazon Q Business | AI assistant over your company's data | Bedrock (you build the app yourself) |
| SageMaker | Full ML platform — build, train, deploy | Bedrock (no model training, just access FMs) |
| Rekognition | Pre-trained vision model, no training needed | SageMaker (needs training) |
Exam tip
If a scenario describes a business that wants to add AI without any ML expertise or model training, the answer is almost always Bedrock or a managed AI service (Rekognition, Comprehend, etc.).
If they want to train their own model on proprietary data, it's SageMaker.