Free AWS AI Practitioner (AIF-C01) Practice Test: 65 Questions with Explanations [2026]

An app lets people find photos by describing them in words, using embeddings and nearest-neighbor lookups over millions of images. Which AWS service can act as the vector database?

A news app compares candidate LLMs on how well they summarize articles against human-written summaries. Which metric fits?

A supply-chain analyst must show executives how each input, such as season or price, shifts a demand model's predictions. Which visualization helps?

During fine-tuning on Amazon Bedrock, a job's loss swings wildly and fails to settle. Which hyperparameter should the team lower first?

A designer's text-to-image model keeps returning generic pictures. Which prompt additions typically improve the result?

Teams edit prompts frequently and need version history, side-by-side comparison and rollback. Which Amazon Bedrock capability fits?

A workflow first asks a model to extract facts from a report, then feeds those facts into a second prompt that drafts a summary. Which technique is this?

A math-help app already knows each student's grade level. Which is the least-effort way to make answers fit that level?

Traffic to a company's customized Bedrock model is predictable and constant around the clock. Which purchasing choice keeps cost lowest?

A bank's fraud alerts are reviewed by analysts, and too many alerts turn out to be legitimate. Which metric should the team improve to cut these false alarms?

A company wants its model to consistently use brand-specific terminology and a fixed output format across all tasks, and it has thousands of labeled examples. Which technique fits best?

A retailer building a recommendation model wants to reduce blind spots that could disadvantage some customers. Which practice helps?

A team designs what goes into each model request: system rules, user history, retrieved articles and tool results. What is this discipline called?

Which TWO content categories can Amazon Bedrock Guardrails content filters detect?

A utility scores 20 million meter readings once every night. Which type of inference fits?

A company wants only its analytics team to invoke certain Amazon Bedrock models. Which solution fits?

A company piloted two models and wants to know which one better serves its goal of cutting handling time. Which evaluation approach is best?

To check a loan model for unfair treatment, a team changes only an applicant's gender in test records and compares the decisions. Which technique is this?

A developer repeatedly rewrites instructions, adds examples and sets context to get better outputs from a model. What is this practice called?

A news app's RAG assistant occasionally drifts into election commentary that the company wants to avoid. Which Amazon Bedrock Guardrails capability blocks a subject?

A telecom wants its chatbot to sound like its best agents and has 200 excellent chat transcripts. Which approach fits?

Models turn words and concepts into lists of numbers that capture meaning. What is the term for these lists?

A shipping company wants a Titan-based assistant to answer from its private manifests. Which solution adds this context without training?

A chatbot's Amazon Bedrock bill rose after developers began sending each user's full chat history with every request. Why?

A company customizes models on Amazon Bedrock and wants the customization artifacts encrypted with a key it creates and controls. Which service fits?

A startup needs a low-cost Amazon Nova model that accepts text and images and works in multiple languages. Which model fits?

A loan model trained mostly on data from urban customers performs poorly for rural applicants. Which type of bias is most likely at work?

Which TWO practices should a team adopt when moving to MLOps?

A bottling plant uses cameras and AI to spot cracked bottles on its line. Which field of AI is this?

A bank's regulator requires that customer data used for AI stay inside a specific country. Which data governance concept addresses this?

A product team wants to try prompts and inference settings against several Amazon Bedrock models before writing any application code. Which capability helps?

A furniture retailer wants shoppers to search by uploading a photo or typing a phrase and get matching products. Which kind of model creates the vectors for this search?

An ad agency wants to generate short promotional video clips from text descriptions with a managed Amazon Bedrock model. Which Amazon Nova model fits?

An insurer has past claims labeled 'fraud' or 'legitimate' and wants a model that labels new claims. Which ML approach fits?

Which of the following is a family of foundation models created by AWS and available through Amazon Bedrock?

Which TWO practices generally improve the quality of a prompt?

A firm wants better answers from a model on Amazon Bedrock at the lowest possible cost. What should it try first?

A prompt includes three sample question-and-answer pairs before the real question. What type of prompting is this?

During which foundation model lifecycle step does a team adapt a pre-trained model by training it further on a few hundred labeled examples for one job?

An insurance company's Amazon Bedrock app runs in a VPC with no internet gateway. How can the app still reach Bedrock privately?

A game studio's image generator sometimes produces characters nearly identical to a rival's copyrighted characters. Which responsible AI concern does this raise?

An HR team's policy chatbot must reflect updated handbook pages within minutes, without retraining. Which approach fits?

A bank's ML model flags loan applications it is unsure about. The bank wants human underwriters to review low-confidence predictions before decisions are sent. Which service fits?

A new colleague asks what an LLM is. Which explanation is accurate?

Auditors ask a bank to prove where each record in its ML training set came from and how it was transformed. Which practice provides this?

A company earns an ISO certification for its AI management system. What does this show?

A team compares embedding models and notices that one uses 3,072 dimensions and another 384. What is the general trade-off of using more dimensions?

A game studio wants an AI opponent that improves by playing many matches and earning points for winning. Which learning approach fits?

Running the same prompt three times gives three differently worded answers from an LLM. Which characteristic of generative AI does this reflect?

A chatbot confidently cites a court case that does not exist. What is this behavior called?

A clinic launches a chatbot that answers patient questions. It wants safeguards that block diagnoses and harmful content in model responses. Which AWS capability fits?

Before archiving stadium camera footage, a team needs to find faces so they can be blurred. Which AWS service detects faces in video?

After customizing a foundation model in Amazon Bedrock, a team must supply a validation dataset. Where should the dataset be stored so Bedrock can read it?

A restaurant chain's marketing analysts, with no coding or ML background, want to forecast weekly sales from internal and external data. Which solution fits?

A recycling plant's model labels each conveyor image as plastic or not plastic. The team wants the overall share of images it labeled correctly. Which metric should it use?

A legal department wants long court rulings condensed into short overviews for busy partners. Which generative AI capability fits?

Before launch, a security team writes deliberately hostile prompts to see whether its chatbot can be tricked, and then adds safeguards for the weaknesses it finds. Which prompting approach is this?

A team runs an AWS Lambda function that calls Amazon Bedrock to summarize support tickets. A developer suggests putting an IAM user's long-term access keys in the function's environment variables. What is the most secure way to give the function permission to invoke the model?

A machine learning lead explains why modern large language models handle long passages better than older sequence models. Which transformer mechanism lets the model weigh how relevant every token is to every other token in the input?

A travel booking site receives hotel reviews written in many languages, and its support staff read only English. Which AWS service should the site use to translate the reviews into English?

A company wants to automatically find personally identifiable information in the Amazon S3 buckets that hold its ML training data. Which service should it use?

An ML team's pipelines rely on outdated libraries, undocumented scripts and quick fixes, so every change is risky. What is this accumulated burden called in MLOps?

A company is drafting an AI governance framework to build trust in its AI systems. Which element belongs in it?

A team trained a model that sorts leaf photos into 12 plant species. It wants a view showing, for each species, how often the model was right or confused it with another species. Which tool provides this?

A travel agency's Amazon Bedrock chatbot must remember that a customer said earlier, "I'm traveling with a toddler." How can the LLM use the earlier messages?