AIF-C01 practice exam and study guide
AWS Certified AI Practitioner
Timed practice exams for AIF-C01 on the real exam's 90 minute limit. Each practice exam serves the same number of questions as the real exam. Every question is original and every option is explained. Your score is broken down by the domains AWS publishes, so you see what to study next. The question bank holds hundreds of practice questions.
Exam facts
- Questions: 65 [1]
- Passing score: 700 on a scaled score of 100 to 1,000 [1]
- Time limit: 90 minutes [1]
- Exam languages (real exam): Arabic, English, French (France), Japanese, Korean, Portuguese (Brazil), Spanish (Latin America), Spanish (Spain), Simplified Chinese, Traditional Chinese [1]
- Delivery: Pearson VUE testing center or online proctored exam [1]
Tokens: 1 for $4.99, 5 for $14.99, 10 for $24.99 or 25 for $49.99. One token runs a full timed AIF-C01 practice exam, and every shorter session on that exam is then free. The Unlimited pass is $149.99 for a year of every practice session on every exam.
Exam domains and weights
AWS publishes these domains and objectives for AIF-C01, quoted here as AWS words them. [2]
Domain 1: Fundamentals of AI and ML, 20% of the exam
- "Task Statement 1.1: Explain basic AI concepts and terminologies."
- "Task Statement 1.2: Identify practical use cases for AI."
- "Task Statement 1.3: Describe the AI/ML development lifecycle."
Domain 2: Fundamentals of GenAI, 24% of the exam
- "Task Statement 2.1: Explain the basic concepts of generative AI (GenAI)."
- "Task Statement 2.2: Understand the capabilities and limitations of GenAI for solving business problems."
- "Task Statement 2.3: Describe AWS infrastructure and technologies for building GenAI applications."
Domain 3: Applications of Foundation Models, 28% of the exam
- "Task Statement 3.1: Describe design considerations for applications that use foundation models (FMs)."
- "Task Statement 3.2: Choose effective prompt engineering techniques."
- "Task Statement 3.3: Describe the training and fine-tuning process for FMs."
- "Task Statement 3.4: Describe methods to evaluate FM performance."
Domain 4: Guidelines for Responsible AI, 14% of the exam
- "Task Statement 4.1: Explain the development of AI systems that are responsible."
- "Task Statement 4.2: Recognize the importance of transparent and explainable models."
Domain 5: Security, Compliance, and Governance for AI Solutions, 14% of the exam
- "Task Statement 5.1: Explain methods to secure AI systems."
- "Task Statement 5.2: Recognize governance and compliance regulations for AI systems."
Sample questions
Three questions from the AIF-C01 practice bank, with the answer and why each option is right or wrong.
Question 1
A team builds a model that reads support messages one word at a time. In the text "The parcel left Lisbon on Monday. By Friday it was still not there," the model must recall Lisbon several sentences later. A standard recurrent neural network (RNN) keeps losing that context. Which model variant should the team use?
- A. Long short-term memory (LSTM) network
- B. Convolutional neural network (CNN)
- C. Feedforward neural network
- D. Feedforward neural network with more hidden layers
Answer: A. A standard RNN can only use the immediate past input. LSTM networks add gated memory cells so the model can keep context across a longer span of a sequence. [3]
- A. Right. An LSTM adds memory cells to the hidden layer, each controlled by input, output and forget gates, so the network can keep useful information over a longer timeline.
- B. Wrong. A CNN is designed to extract spatial information from images and video through convolutional and pooling layers. It is not built to carry context across a word sequence.
- C. Wrong. A feedforward network passes information from one end to the other and cannot remember an input it has already processed. It would lose Lisbon sooner than the RNN does.
- D. Wrong. Extra hidden layers do not add memory. A feedforward network of any depth still forgets each earlier input once it moves on.
Question 2
A company runs an old desktop billing application that has no API. It wants an AI agent to work the application the way a clerk does: read the screen, click through menus and type values into fields. Which agent pattern meets this need?
- A. Basic reasoning agent
- B. Coding agent
- C. Computer-use agent
- D. Speech and voice agent
Answer: C. When a system offers no API, a computer-use agent can act through its user interface, combining LLM reasoning with a visual language model that reads the screen. [4]
- A. Wrong. This pattern cannot operate the screens, menus and fields of an application.
- B. Wrong. The billing application has no API and must be worked through its screens. A coding agent is not the pattern that observes a screen and sends input events to an application.
- C. Right. Computer-use agents control digital environments such as browsers, terminals and applications. A visual language model observes the screen and the agent performs actions the way a human would.
- D. Wrong. A voice agent works through spoken conversation. The billing application needs screen reading and input events.
Question 3
A company's generative AI application handles retrieval, summarization and formatting in one large prompt inside one application component. Every small prompt change to improve one part of the output breaks other parts, and the component is hard to test. Which design should the company move to?
- A. A larger model that can handle the whole task in one prompt.
- B. One longer prompt that repeats every instruction twice.
- C. A compound AI system that splits the task into smaller, loosely coupled steps, each run as its own microservice.
- D. A single application component with more memory and CPU.
Answer: C. A single component that does everything is a known anti-pattern for generative AI in production. Break the work into focused steps that can fail, change and scale separately. [5]
- A. Wrong. A bigger model keeps the monolithic design, so one change can still break every part of the output.
- B. Wrong. Repeating instructions keeps everything in one brittle component and does not make it easier to test.
- C. Right. Decomposing a monolith into a compound AI system, a chain of smaller discrete steps, lets each component be developed, deployed, scaled and monitored on its own.
- D. Wrong. More resources do not fix a design that is brittle, hard to test and risky to update.
Practice for AIF-C01 on ExamSens.ai
- Set up a timed practice exam for AIF-C01. One token runs a full exam, then every shorter session on that exam is free.
- Study the AIF-C01 flashcards, one card for each rule in the question bank.
- The AIF-C01 study guide: 5 tokens to read it online and download the PDF, with 1 free exam token for any practice exam.
- Practice tests on Udemy: Part 1 | Part 2
- Token packs and the Unlimited pass.
ExamSens.ai is not authorized, endorsed or sponsored by Amazon Web Services.
References
[1] https://aws.amazon.com/certification/certified-ai-practitioner/
[3] https://aws.amazon.com/what-is/recurrent-neural-network/