Domain 1 Mock Exam — 50 Questions (Lessons 1–3 cumulative)

What is the broadest category that contains Machine Learning, Deep Learning, and Generative AI?

Which inferencing type processes many inputs together on a schedule, nobody waiting live?

Which learning type uses labeled data, where every example has the correct answer attached?

Which AWS service detects sentiment of thousands of customer reviews?

Which technique groups data points by similarity with no predefined labels?

Which metric measures the fraction of ALL predictions that were correct?

What is the term for a deployed model's performance degrading as live data diverges from training data?

Which deployment method means AWS manages the infrastructure and you just send/receive requests?

Which AWS offering gives access to multiple FMs from different providers through one API, plus Knowledge Bases and Guardrails?

Which is a business metric rather than a model/technical metric?

Which describes an 'algorithm' as opposed to a 'model'?

A system spotting defective products via camera images is an example of which AI capability?

Using an already-trained open-source model largely as-is is an example of which FM source?

A task requiring an exact, guaranteed answer every time (e.g., payroll tax) is generally a poor ML fit because:

Which term describes treating ML systems as an ongoing, repeatable, managed process?

A logistics company collects GPS pings every 5s from trucks, historical routes labeled 'on-time'/'delayed', wants instant delay alerts. Correct (1) data type, (2) learning type, (3) inferencing type?

A healthcare company must fully explain/audit every factor behind a diagnosis-support prediction for regulatory compliance. Better fit?

Airport security screening missing a genuine threat (false negative) is far more dangerous than a false alarm. Prioritize?

Which are valid reasons an AI/ML solution might NOT be appropriate for a business problem?

A company wants to convert English text into natural-sounding speech for an IVR system, no custom training. Best fit?

Which AWS service detects data quality/model quality/bias drift by comparing live data against a training baseline?

A recommendation model has strong precision/recall but costs $5/prediction while generating only $1 additional revenue per recommendation. Conclusion?

A company records calls in French, needs an English sentiment report. Order the AWS services correctly.

A retail company wants a chatbot that understands questions and routes them to the right department, no custom model. Best fit?

Match each MLOps concept to its definition.

A system is given the goal 'resolve this billing dispute' and independently checks the database, calculates a refund, issues it, and emails confirmation with no human approval. Best described as:

Which correctly describe 'bias' in an AI/ML model?

Which AWS service is specifically designed to help detect bias in datasets and models?

A model performs poorly on both training data and new data, never having captured the underlying pattern. This is:

Match each inferencing type to its scenario.

Which algorithm builds many small decision trees in sequence, each correcting prior errors?

Thousands of raw scanned PDF documents with no predefined rows/columns — what data type?

Predicting next month's total sales revenue using 5 years of monthly historical data is best described as:

Which AWS tool is a full AI-powered development environment (agentic IDE), distinct from Amazon Q's chat assistant?

A model reports 96% accuracy where 95% of cases are the majority class. Using the naive-baseline sanity check, conclusion?

A bank's model has 99.5% accuracy detecting fraud but only catches 10% of actual fraud (low recall), with high precision. Missing fraud costs far more than a false alarm. What should the bank conclude?

A company is deciding on a foundation model for an open-ended internal chatbot summarizing arbitrary policy documents. Which support using a FOUNDATION MODEL?

Order these MLOps-related events chronologically for a deployed fraud model experiencing drift.

Model X: 85% precision, 80% recall, $0.10/prediction. Model Y: 95% precision, 92% recall, $3.00/prediction. Fraud loss prevented per catch = $2.50. Which delivers better ROI at high volume?

Which statements accurately distinguish 'asynchronous inferencing' from 'batch inferencing'?

A model trained on complaint text only from English-speaking urban call centers performs noticeably worse for rural, non-English-dominant regions even after proper translation. Most likely root cause?

Match each scenario to the concept/metric it should prioritize.

A company pulls a pre-trained open-source LLM and hosts it themselves on their own SageMaker AI infrastructure, managing all scaling/maintenance in-house. Correct (1) FM source, (2) deployment method?

Which are true about Amazon Bedrock?

A model finishes training with excellent evaluation metrics, but no monitoring, logging, or rollback plan is set up before going live. Most accurate assessment?

Order these from BROADEST concept to a single concrete example, by increasing specificity.

A chatbot is evaluated with ROUGE/BLEU against reference answers and scores low, even though human reviewers rate replies as excellent but differently phrased. Most likely explanation?

Which are accurate about the relationship between technical (model) metrics and business metrics?

A company builds a RAG-style internal search assistant using Bedrock. Order these steps correctly.

Why might a company choose a traditional ML classification model over a Bedrock foundation model for real-time credit card fraud detection at massive volume?