AIF-C01 sample questions with answers
The other 87 questions are in the practice tool, with answers revealed as you go.
Start 20 questionsFundamentals of AI and ML
- A manufacturer streams vibration readings from its pumps. Nearly all readings are normal, and the team wants to flag rare readings that deviate sharply from typical behavior. It has no labeled examples of failures. Which kind of task is this?
- A manufacturer wants a model that inspects photos of circuit boards on a production line and flags visible solder defects. Which neural network architecture is best suited to this task?
- A clinic has only a few hundred labeled X-ray images for a model that flags a rare condition. Which technique creates more varied training examples without collecting new scans?
- A credit-scoring model's production inputs have gradually shifted away from the sample it was trained on, so its predictions are becoming less reliable. What is this phenomenon called?
Fundamentals of generative AI
- A developer wants generated text to draw only from the smallest group of next-token candidates whose combined probability reaches a chosen threshold, so rare words are excluded while the output still varies. Which inference parameter does this?
- Which transformer component lets each token weigh every other token in the sequence when it builds its representation?
Applications of foundation models
- A company wants to serve a support chatbot at high volume with lower per-request cost. It trains a compact model to reproduce the answers of a much larger model, and the compact model then handles the traffic. What is this technique called?
- A team building semantic search over product manuals needs a managed AWS store that can hold embedding vectors and run k-nearest-neighbor queries. Which service should it choose?
- A media company runs a steady, high-volume summarization job through a foundation model around the clock and needs predictable latency at a lower unit cost. Which Amazon Bedrock option should it evaluate?
- A product team wants to add summarization to its app using ready-made models, while a separate research group wants to build and train its own forecasting models on proprietary data. Which pairing of AWS services matches those needs?
Guidelines for responsible AI
Security, compliance and governance for AI
- A clinic runs a patient-intake assistant on AWS and must meet its HIPAA obligations. Which task remains the clinic's own responsibility rather than AWS's?
- A law firm sends privileged client memos through Amazon Bedrock prompts. What assurance matters most for whether those memos could be used to improve the foundation models?