Amazon AWS Certified Machine Learning – Specialty (MLS-C01) Exam Dumps and Practice Test Questions Set 8 Q141-160

Visit here for our full Amazon AWS Certified Machine Learning – Specialty exam dumps and practice test questions.

Question 141

A company wants to deploy multiple ML models on a single endpoint, loading models on demand to save memory. Which feature should they use?

A) SageMaker Multi-Model Endpoints
B) SageMaker Asynchronous Inference
C) ECS Auto Scaling
D) EC2 Spot Instances

Answer: A

Explanation:

SageMaker Multi-Model Endpoints allow multiple ML models to be hosted on a single endpoint. These endpoints load models dynamically from Amazon S3 into memory only when a request for that specific model is made. This approach minimizes memory usage because only the models that are actively being queried consume resources, making it ideal for serving hundreds or thousands of models without requiring a separate endpoint for each one. Multi-Model Endpoints also support versioning and scaling to handle varying traffic loads efficiently.

SageMaker Asynchronous Inference is designed for workloads where predictions are long-running or large in size. It allows users to submit inference requests and retrieve results later. While it supports batch-like processing, it does not provide a mechanism to dynamically host multiple models on the same endpoint, and memory savings through on-demand model loading are not a feature of this service.

ECS Auto Scaling is a feature of the Elastic Container Service that allows containers to scale based on metrics such as CPU or memory usage. While it helps in managing infrastructure efficiently, it does not inherently provide ML model hosting capabilities. Managing multiple models in ECS requires manual orchestration, and scaling does not address the memory overhead of loading numerous models simultaneously.

EC2 Spot Instances offer a cost-effective way to run workloads on unused EC2 capacity. Spot Instances can reduce compute costs but do not provide ML-specific features like model hosting, dynamic loading, or automated scaling of endpoints. Users would need to manually deploy and manage models across multiple instances, which adds operational complexity.

Multi-Model Endpoints are the correct solution because they are purpose-built to host multiple models on a single endpoint, dynamically loading only the required models to conserve memory. Unlike the other options, Multi-Model Endpoints combine cost efficiency, scalability, and operational simplicity specifically for ML inference workloads.

Question 142

A company wants to forecast product demand for thousands of items using historical data and related datasets. Which AWS service should they use?

A) Amazon Forecast
B) SageMaker Autopilot
C) AWS Lambda
D) Lookout for Metrics

Answer: A

Explanation:

Amazon Forecast is a fully managed service designed for time-series forecasting. It can automatically process historical data and related datasets, such as promotions, holidays, and seasonality, to generate accurate predictions. Forecast supports modeling thousands of items simultaneously and can provide forecasts at item, category, or location levels. The service abstracts the complexities of building, training, and tuning machine learning models, making it suitable for demand forecasting at scale.

SageMaker Autopilot is an automated machine learning service that simplifies creating and deploying models. While it can be used for regression or classification tasks, it is not specialized for time-series forecasting. Handling temporal dependencies and related datasets in Autopilot requires manual configuration, and the resulting models may not provide the same accuracy or scalability as Amazon Forecast.

AWS Lambda provides serverless compute capabilities to run code in response to events. It is useful for data processing or triggering workflows but does not include time-series modeling or automated forecasting functionality. Users would need to build custom forecasting pipelines using external libraries, which increases complexity.

Lookout for Metrics is focused on anomaly detection. It monitors metrics and identifies deviations from expected behavior, helping detect unexpected spikes or drops in demand. However, it is not designed to generate future forecasts based on historical trends, so it cannot meet the requirement for proactive demand planning.

Amazon Forecast is the ideal choice because it is specifically built for large-scale demand forecasting. It automates feature engineering, model selection, and training, supports multiple items and related datasets, and provides accurate future predictions, making it more suitable than general-purpose ML or compute services.

Question 143

A startup wants to perform multi-node GPU training for a large deep learning model with minimal setup. Which service is most suitable?

A) SageMaker Distributed Training
B) Lambda
C) AWS Glue
D) Rekognition

Answer: A

Explanation:

SageMaker Distributed Training is designed for large-scale deep learning workloads that require multiple GPUs across multiple nodes. It handles the orchestration of data parallelism, model parallelism, and gradient aggregation automatically, reducing the operational complexity of distributed training. Developers can scale up training jobs with minimal configuration and leverage high-performance GPU instances to accelerate model convergence.

AWS Lambda is a serverless compute service designed for short-lived, event-driven tasks. Lambda does not support GPU acceleration or long-running training processes, and the execution time limits make it unsuitable for large-scale deep learning. Attempting to use Lambda for multi-node GPU training would not be feasible.

AWS Glue is a fully managed ETL service used for data preparation and transformation. While it can handle large datasets, it is not designed for training deep learning models or managing GPU clusters. Using Glue for model training would require extensive customization and additional infrastructure.

Rekognition is an AI service focused on computer vision tasks such as image and video analysis. It does not provide functionality for training custom deep learning models across multiple nodes or GPUs. It is only suitable for pre-trained vision applications.

Distributed Training is the correct choice because it provides the necessary infrastructure, orchestration, and scalability for multi-node GPU training. It minimizes setup effort while enabling efficient utilization of compute resources, which is critical for large deep learning models.

Question 144

A company wants to deploy RL models to edge devices with version control and automatic updates. Which service should they use?

A) SageMaker Edge Manager
B) SageMaker Processing
C) AWS Batch
D) AWS Glue

Answer: A

Explanation:

SageMaker Edge Manager is a service specifically designed to deploy, monitor, and manage ML models on edge devices. It supports versioning, enabling teams to update models without downtime, and provides automatic updates to edge devices. Edge Manager also tracks model performance and resource usage on devices, ensuring consistent operation in production environments.

SageMaker Processing is intended for preprocessing and postprocessing of data during ML workflows. It allows running scripts for data transformation and feature engineering but does not facilitate deployment or monitoring of models on edge devices. It is purely for data handling, not edge inference.

AWS Batch is a fully managed service to execute batch computing jobs at scale. While it can run containerized workloads, it does not provide edge deployment, model versioning, or automatic update capabilities, making it unsuitable for RL model distribution on edge devices.

AWS Glue is an ETL service for extracting, transforming, and loading data. It has no functionality for ML deployment or edge device management, so it cannot address the requirements of versioned RL model deployment or monitoring.

Edge Manager is the correct solution because it provides all the features needed to manage reinforcement learning models on edge devices. It ensures seamless updates, version control, and monitoring, making it purpose-built for production RL deployments.

Question 145

A team wants to track ML experiments including datasets, hyperparameters, and metrics with visual comparison. Which service is best?

A) SageMaker Experiments
B) SageMaker Data Wrangler
C) SageMaker Canvas
D) SageMaker Edge Manager

Answer: A

Explanation:

SageMaker Experiments is designed to help data scientists organize, track, and compare ML experiments. It stores datasets, hyperparameters, and metrics from each training run and provides a visual interface to compare experiments. This allows teams to identify the most effective models and configurations systematically, supporting reproducibility and collaboration across projects.

SageMaker Data Wrangler is a feature engineering tool for preprocessing and transforming data. While it enables visual data cleaning and transformation, it does not track experiments, hyperparameters, or training metrics, so it cannot fulfill the requirements for experiment management.

SageMaker Canvas is a no-code ML tool for business analysts to generate predictions from data. It simplifies model building but does not provide experiment tracking, hyperparameter management, or detailed metrics comparison. It is more suitable for non-technical users rather than data science teams focused on experimentation.

SageMaker Edge Manager manages model deployment and monitoring on edge devices. It provides versioning, updates, and device-level monitoring but does not track ML experiments or compare metrics, so it cannot meet the requirements of experiment management.

SageMaker Experiments is the correct choice because it is explicitly built to track datasets, hyperparameters, and metrics across multiple runs. Its visual comparison capabilities allow teams to systematically analyze and select the best performing models, making it essential for organized ML experimentation.

Question 146

A company wants to monitor deployed ML models for bias and explainability. Which service should they use?

A) SageMaker Clarify
B) SageMaker Model Monitor
C) CloudWatch Metrics
D) AWS Glue

Answer: A

Explanation:

SageMaker Clarify is designed to provide both bias detection and explainability for machine learning models. It can analyze both training data and model outputs to identify features that may contribute to biased predictions, and it can generate detailed reports highlighting potential fairness issues. Clarify also provides explainability metrics such as SHAP values to show which features most influence model predictions, making it valuable for regulated industries and transparent AI workflows.

SageMaker Model Monitor is intended for monitoring the ongoing quality and performance of deployed models. It tracks data drift, feature distributions, and model accuracy over time. While this service is important for operational ML monitoring, it does not provide insights into bias or model explainability and therefore cannot fully address fairness concerns.

CloudWatch Metrics is a general monitoring tool for AWS resources. It is excellent for tracking infrastructure metrics such as CPU usage, memory, latency, and request counts. However, it is not tailored for machine learning analysis, and it does not provide any functionality to assess model bias or explain predictions.

AWS Glue is a managed ETL service used for extracting, transforming, and loading data at scale. While it is highly useful for preprocessing and preparing datasets, it does not offer ML model monitoring, bias detection, or explainability features. Its focus is on data pipelines rather than model assessment.

The correct choice is SageMaker Clarify because it is purpose-built for detecting bias and evaluating explainability in machine learning workflows. Unlike the other options, Clarify integrates seamlessly with SageMaker pipelines, supports both pre-training and post-training analysis, and provides detailed reports that help data scientists and compliance teams ensure that models are fair, interpretable, and accountable.

Question 147

A company wants to preprocess large-scale image data in a managed distributed environment integrated with S3. Which service should they use?

A) SageMaker Processing
B) Amazon EMR
C) AWS Glue
D) EC2 Auto Scaling

Answer: A

Explanation:

SageMaker Processing allows companies to run distributed data preprocessing, feature engineering, and model evaluation jobs. It provides fully managed compute resources and integrates natively with Amazon S3, enabling seamless loading and saving of large datasets. For image data, Processing can handle tasks such as resizing, normalization, augmentation, and conversion in a scalable way without manual cluster management.

Amazon EMR is a managed big data platform for processing large datasets using frameworks like Spark, Hadoop, and Presto. While EMR can be used for distributed data processing, it is more general-purpose and requires extra setup for handling machine learning-specific preprocessing pipelines. It lacks the tight integration with SageMaker ML workflows that Processing provides.

AWS Glue is a serverless ETL service used for preparing structured or semi-structured data. Glue excels at data cleaning, transformation, and cataloging for analytics workloads. It is not optimized for image data preprocessing and does not provide ML-specific distributed processing capabilities out of the box.

EC2 Auto Scaling is designed to automatically adjust the number of EC2 instances to meet demand. While it provides scalable compute infrastructure, it requires manual orchestration of preprocessing jobs and does not offer managed distributed ML workflows or native S3 integration for ML pipelines.

SageMaker Processing is the best option because it combines managed distributed compute, seamless S3 integration, and ML workflow support. It allows teams to preprocess large-scale image datasets efficiently, without the operational overhead of managing clusters or pipelines manually.

Question 148

A company wants to run large-scale batch inference without requiring real-time predictions. Which service is most suitable?

A) SageMaker Batch Transform
B) SageMaker Real-Time Inference
C) SageMaker Serverless Inference
D) AWS Lambda

Answer: A

Explanation:

SageMaker Batch Transform is designed for asynchronous batch inference on large datasets. It allows companies to run high-volume predictions efficiently, without the need for maintaining persistent endpoints. It handles model loading, parallel processing, and scaling automatically, making it ideal for scenarios where predictions are needed for large datasets rather than in real-time.

SageMaker Real-Time Inference provides low-latency, online predictions using dedicated endpoints. It is optimized for scenarios where each request requires an immediate response. For large-scale batch jobs that do not require instant predictions, this approach is inefficient and can be more costly.

SageMaker Serverless Inference is a newer option for real-time inference without managing infrastructure. While it reduces operational overhead and scales automatically, it is still intended for online requests rather than high-volume batch workloads.

AWS Lambda is a serverless compute service capable of running arbitrary functions in response to events. While Lambda can process small batches, it is limited by execution duration and memory, making it unsuitable for large-scale inference tasks.

Batch Transform is the most appropriate choice because it is optimized for high-volume, asynchronous predictions, minimizes operational overhead, and integrates seamlessly with SageMaker models and S3 data storage.

Question 149

A company wants to detect anomalies in business metrics automatically. Which service should they choose?

A) Lookout for Metrics
B) Amazon Forecast
C) SageMaker Autopilot
D) AWS Lambda

Answer: A

Explanation:

Lookout for Metrics is an automated anomaly detection service designed to monitor time-series data and business metrics. It uses machine learning to detect anomalies in patterns, accounting for seasonality and trends. The service can send alerts when unexpected deviations occur, making it highly suitable for proactive monitoring of KPIs and operational metrics.

Amazon Forecast is primarily used for predicting future trends and demand using time-series data. While it can generate forecasts for business metrics, it is not designed to detect anomalies in historical data and will not alert teams about unexpected deviations.

SageMaker Autopilot automates the process of building, training, and deploying machine learning models. It is useful for general-purpose ML but does not provide specialized anomaly detection or metric monitoring capabilities out of the box.

AWS Lambda provides serverless compute capabilities and can be used to run custom scripts for anomaly detection. However, implementing an automated, scalable solution would require significant manual development and integration.

Lookout for Metrics is the correct choice because it is purpose-built for anomaly detection, requires minimal setup, handles time-series data efficiently, and automatically identifies deviations in business metrics without additional ML model development.

Question 150

A healthcare startup wants to label sensitive images securely with HIPAA compliance. Which service should they use?

A) SageMaker Ground Truth Private Workforce
B) Mechanical Turk
C) AWS Batch
D) Rekognition Custom Labels

Answer: A

Explanation:

SageMaker Ground Truth Private Workforce allows companies to securely label data within a private, VPC-isolated environment. It supports HIPAA-compliant workflows, audit logging, and role-based access, ensuring sensitive healthcare images are handled securely. Human labelers are vetted and controlled, providing compliance and confidentiality.

Mechanical Turk is a public crowdsourcing platform. While it can label images at scale, it does not provide privacy isolation or HIPAA compliance, making it unsuitable for sensitive healthcare data.

AWS Batch is a service for running large-scale compute jobs. It is useful for processing tasks but does not provide human labeling or privacy compliance features.

Rekognition Custom Labels automates image labeling using computer vision models. While it can identify features and categorize images, it does not provide private human labeling with HIPAA compliance, which is often required for sensitive healthcare datasets.

Ground Truth Private Workforce is the correct choice because it ensures HIPAA-compliant, private, and auditable labeling of sensitive images, which is crucial for healthcare applications.

Question 151

A startup wants to deploy thousands of small ML models efficiently on a single endpoint, loading models dynamically. Which feature should they use?

A) SageMaker Multi-Model Endpoints
B) SageMaker Asynchronous Inference
C) ECS Auto Scaling
D) EC2 Spot Instances

Answer: A

Explanation:

SageMaker Multi-Model Endpoints allow multiple models to be hosted on a single endpoint while dynamically loading models from Amazon S3 only when requests are made for them. This design reduces memory consumption and improves cost efficiency because it avoids keeping all models in memory simultaneously. It is particularly useful for organizations that need to serve a very large number of smaller models without creating an endpoint for each one.

SageMaker Asynchronous Inference is intended for handling requests that require long processing times. It queues requests and returns predictions once completed, but it does not manage multiple models dynamically on a single endpoint. While it supports large workloads, it focuses more on request management than on memory-efficient model hosting.

ECS Auto Scaling allows containerized applications to automatically scale based on demand. While this could theoretically help scale multiple inference containers, it does not provide native mechanisms to dynamically load ML models on demand or consolidate many models on a single endpoint. Manual orchestration is required to handle model deployment efficiently.

EC2 Spot Instances are cost-effective compute resources that can be interrupted when AWS needs the capacity back. They are useful for training or batch processing but do not offer a managed solution for hosting multiple models dynamically. Using Spot Instances for this purpose would require significant manual orchestration and management.

The correct answer is SageMaker Multi-Model Endpoints because it is specifically designed to host many models on a single endpoint, load models dynamically from S3, and conserve memory. It simplifies deployment and ensures scalable, cost-effective hosting without the need for custom orchestration.

Question 152

A company wants to forecast deliveries across multiple locations using historical and related datasets. Which service should they use?

A) Amazon Forecast
B) SageMaker Autopilot
C) AWS Lambda
D) Lookout for Metrics

Answer: A

Explanation:

Amazon Forecast is a fully managed service for time-series forecasting that leverages historical data and related datasets to predict future trends. It automatically handles data preprocessing, model selection, hyperparameter tuning, and supports multiple items, such as different locations or products. It also allows users to incorporate holidays, promotions, and other factors that can impact forecasts, making it well-suited for delivery predictions across multiple sites.

SageMaker Autopilot automates end-to-end machine learning workflows, creating and tuning models automatically. While Autopilot can build predictive models from datasets, it is not specialized for time-series forecasting. Forecast provides optimized, purpose-built algorithms that outperform general ML models for this type of application.

AWS Lambda is a serverless compute service that executes code in response to events. It is not an ML service and cannot perform forecasting by itself. While Lambda could be used to orchestrate parts of a forecasting pipeline, it does not provide modeling, training, or evaluation capabilities.

Lookout for Metrics automatically detects anomalies in time-series or business data. It helps identify unexpected changes in metrics but does not provide forecasting or predictive capabilities. It is useful for monitoring but not for planning or predicting future delivery volumes.

Amazon Forecast is the correct choice because it is explicitly designed for multi-location, time-series forecasting using historical and related datasets. It offers automated model selection, handles multiple items and holidays, and provides accurate, scalable predictions.

Question 153

A startup wants to run multi-node GPU training for a large NLP model. Which service is most suitable?

A) SageMaker Distributed Training
B) Lambda
C) AWS Glue
D) Rekognition

Answer: A

Explanation:

SageMaker Distributed Training simplifies the process of training large ML models across multiple GPU instances. It manages the orchestration of nodes, distributes model weights, and synchronizes gradients automatically. This allows data scientists to scale efficiently for compute-intensive NLP workloads without manually handling parallelization or synchronization.

AWS Lambda provides serverless compute capabilities for short-lived tasks. It does not support GPU-based workloads and is unsuitable for multi-node, high-memory training jobs. While Lambda excels in lightweight event-driven processing, it cannot handle large-scale model training.

AWS Glue is a managed ETL service that focuses on data preparation and transformation. It can orchestrate pipelines and perform batch processing but has no native support for GPU-based training or model distribution. Glue is designed for data workflows rather than training complex ML models.

Amazon Rekognition is an image and video analysis service offering pre-built models for computer vision tasks such as object detection and facial recognition. It does not provide tools for training custom NLP models or managing multi-node GPU workloads.

SageMaker Distributed Training is the correct option because it enables efficient multi-node, multi-GPU training for large NLP models. It automates parallelization and synchronization, which would otherwise require significant manual effort, making it ideal for startups working with complex models.

Question 154

A company wants to deploy RL models to edge devices with version control and automatic updates. Which service should they use?

A) SageMaker Edge Manager
B) SageMaker Processing
C) AWS Batch
D) AWS Glue

Answer: A

Explanation:

SageMaker Edge Manager is specifically designed to manage machine learning models on edge devices. It provides a comprehensive set of tools for packaging, deploying, monitoring, and updating models in distributed environments. This includes support for reinforcement learning (RL) models, which often require frequent updates and monitoring to maintain optimal performance in dynamic conditions. Edge Manager simplifies these processes by providing centralized control over deployed models, making it easier for teams to ensure consistent behavior across multiple devices. It also enables version control, so teams can track which model version is running on each device and roll back or update models as needed.

SageMaker Processing, on the other hand, is primarily focused on data preprocessing, postprocessing, and feature engineering tasks. It is useful for preparing datasets before training or transforming outputs after inference, providing a scalable way to manage data workflows. While Processing can handle tasks such as data cleaning, transformation, and feature extraction, it does not offer functionality for deploying models to edge devices or managing updates in distributed systems. It is primarily concerned with the data side of machine learning rather than operationalizing models in production at the edge.

AWS Batch is a managed service designed for running batch computing workloads at scale. It allows users to schedule and execute large volumes of jobs on EC2 instances efficiently. However, AWS Batch does not include features for deploying, monitoring, or updating machine learning models on edge devices. Its primary use case is high-throughput computing or large-scale data processing, rather than real-time inference or edge deployment. While it is powerful for processing tasks, it is not intended for managing distributed ML models.

AWS Glue is an extract, transform, and load (ETL) service that automates data integration and pipeline creation. Glue is highly effective for moving and transforming data across data stores but does not provide capabilities for deploying or updating machine learning models. Its focus is on data workflows and automation rather than edge device management, so it is not suitable for organizations looking to operationalize ML models on distributed devices.

SageMaker Edge Manager is the correct choice because it provides an end-to-end solution for deploying, updating, and monitoring machine learning models on edge devices. It ensures version control, centralized monitoring, and seamless updates, which are critical for maintaining the performance and reliability of reinforcement learning models and other ML workloads in edge environments. By integrating deployment, versioning, and monitoring in a single managed service, Edge Manager reduces operational complexity and ensures that distributed models remain secure, up-to-date, and consistent across all devices.

Question 155

A team wants to track ML experiments including datasets, hyperparameters, and metrics visually. Which service should they use?

A) SageMaker Experiments
B) SageMaker Data Wrangler
C) SageMaker Canvas
D) SageMaker Edge Manager

Answer: A

Explanation:

SageMaker Experiments is a service designed to help teams systematically track and manage the lifecycle of machine learning experiments. It captures detailed information about each training run, including the datasets used, hyperparameters configured, evaluation metrics, and outputs generated by the model. By organizing this information in a structured manner, Experiments enables teams to visualize and compare results across different runs. This makes it easier to identify which configurations lead to the best-performing models, improving decision-making during model development. Additionally, by maintaining a history of experiments, it supports reproducibility and accountability, which are critical in professional ML workflows and regulated industries.

SageMaker Data Wrangler, by contrast, focuses primarily on data preparation and feature engineering. It provides a visual interface to clean, transform, and explore datasets before feeding them into machine learning models. While Data Wrangler is extremely useful for accelerating data preprocessing tasks and simplifying the creation of feature pipelines, it does not offer the functionality to track experiments, record hyperparameters, or compare model metrics. Its main purpose is to ensure that datasets are ready for training rather than managing the lifecycle of model experimentation.

SageMaker Canvas is a no-code machine learning tool that allows business analysts and users with limited coding experience to build models using a visual interface. It simplifies ML model creation for non-technical users by abstracting the underlying code and algorithm selection. However, Canvas does not provide detailed experiment tracking, hyperparameter tuning, or metric comparison, making it unsuitable for professional ML workflows where rigorous tracking and reproducibility of experiments are required. It is better suited for rapid prototyping or generating insights without deep ML expertise.

SageMaker Edge Manager is focused on managing, deploying, and monitoring models on edge devices. It provides capabilities for version control, model updates, and performance monitoring in edge environments. While Edge Manager is valuable for operationalizing ML models in distributed environments, it is not intended for experiment tracking or visual analysis of model performance during the training and development phases.

SageMaker Experiments is the correct choice because it combines comprehensive experiment tracking, parameter recording, and metric visualization in one service. It allows teams to compare models effectively, identify trends, and determine which configurations perform best, all while ensuring transparency and reproducibility. By integrating with other SageMaker components, it also facilitates a seamless workflow from experiment tracking to model deployment, supporting both individual data scientists and collaborative teams in maintaining structured and accountable ML development processes.

Question 156

A company wants to monitor deployed ML models for bias and explainability. Which service should they choose?

A) SageMaker Clarify
B) SageMaker Model Monitor
C) CloudWatch Metrics
D) AWS Glue

Answer: A

Explanation:

SageMaker Clarify is a service specifically designed to detect and mitigate bias in machine learning models. It can evaluate both pre-training and post-training datasets to identify potential sources of bias, such as unbalanced classes or features that may unintentionally favor certain outcomes. Additionally, Clarify generates explainability reports that highlight which input features contribute most to predictions, making it easier for teams and stakeholders to understand model behavior and ensure fairness.

SageMaker Model Monitor, while related to deployed models, primarily focuses on monitoring model quality over time. It tracks metrics like data drift, feature drift, and model performance degradation, helping detect when a model may require retraining. However, it does not provide detailed bias detection or explainability features, which are essential for evaluating fairness in predictions.

CloudWatch Metrics is a monitoring tool for AWS infrastructure and application metrics. It can collect and track metrics from EC2 instances, SageMaker endpoints, or other AWS resources, and can trigger alerts when thresholds are crossed. While useful for operational monitoring, it does not analyze model predictions for bias or provide explainability reports, making it unsuitable for fairness monitoring in ML.

AWS Glue is an extract, transform, and load (ETL) service used for preparing and moving data between data stores. It is valuable for data preprocessing and cleaning but does not provide any specialized functionality for monitoring model bias or explainability. Its role is purely in data management, not model evaluation.

Given these options, SageMaker Clarify is the correct choice because it directly addresses the need to monitor deployed ML models for both bias and explainability. It provides automated analysis, clear reporting, and integration with SageMaker workflows, enabling teams to maintain fairness and transparency in AI applications.

Question 157

A company wants to preprocess large-scale image data using a managed distributed ML service integrated with S3. Which service is most suitable?

A) SageMaker Processing
B) EMR
C) AWS Glue
D) EC2 Auto Scaling

Answer: A

Explanation:

SageMaker Processing allows users to run large-scale data preprocessing jobs in a managed environment. It integrates seamlessly with S3 for data input and output, supports distributed computing, and can automatically scale the underlying infrastructure. It is particularly useful for image and tabular data transformations before model training, ensuring reproducible and efficient preprocessing pipelines.

Amazon EMR is a general-purpose big data platform for processing large datasets using frameworks like Spark or Hadoop. While powerful for batch processing and analytics, it is not tailored for machine learning preprocessing and may require significant manual configuration for tasks such as image augmentation or feature extraction.

AWS Glue focuses on ETL workflows and is optimized for transforming structured data for analytics pipelines. It lacks built-in support for complex image preprocessing or distributed ML-specific operations, making it less suitable for large-scale ML data preparation.

EC2 Auto Scaling is a feature that dynamically adjusts the number of EC2 instances based on demand. While it can help scale compute resources, it does not provide any prebuilt ML preprocessing tools or integration with S3, meaning users would need to implement custom scripts and orchestration manually.

Considering these factors, SageMaker Processing is the best fit because it combines managed distributed computing, S3 integration, and ML-specific preprocessing tools. This allows teams to efficiently handle large-scale image datasets without manual infrastructure management.

Question 158

A company wants to run large-scale batch inference without requiring real-time predictions. Which service is best?

A) SageMaker Batch Transform
B) SageMaker Real-Time Inference
C) SageMaker Serverless Inference
D) AWS Lambda

Answer: A

Explanation:

SageMaker Batch Transform is explicitly designed for running batch inference on large datasets asynchronously. Users can submit data in bulk from S3, and the service handles distributing computation across instances automatically. Batch Transform is ideal when predictions are needed for large datasets without the need for low-latency responses.

SageMaker Real-Time Inference provides online, low-latency predictions through a persistent endpoint. It is suitable for interactive applications, but it is inefficient and costly for batch scoring large datasets, as the endpoint must remain active during the entire inference process.

SageMaker Serverless Inference is a newer option for infrequent or unpredictable workloads that require online predictions without dedicated endpoints. While it scales automatically, it is still designed for online inference rather than scheduled or bulk batch scoring, limiting its usefulness for large-scale batch tasks.

AWS Lambda is a serverless compute service that can run code in response to events. While flexible, Lambda is limited in memory and runtime duration, making it impractical for high-volume ML inference jobs without significant orchestration and chunking.

Batch Transform is the correct choice because it allows efficient, scalable, and cost-effective large-scale inference. It handles asynchronous processing, integrates directly with S3 for input and output, and removes the need to maintain real-time endpoints for batch workloads.

Question 159

A company wants to detect anomalies in business metrics automatically. Which service should they use?

A) Lookout for Metrics
B) Amazon Forecast
C) SageMaker Autopilot
D) AWS Lambda

Answer: A

Explanation:

Lookout for Metrics is a fully managed AWS service specifically designed for automatically detecting anomalies in time-series data. It uses machine learning to continuously monitor metrics such as sales, revenue, website traffic, or operational KPIs and can identify unusual patterns or unexpected changes. These anomalies could manifest as sudden spikes, unexpected drops, or gradual deviations from expected behavior. One of the key strengths of Lookout for Metrics is that it not only detects these anomalies automatically but also provides root-cause analysis. This helps teams understand the factors contributing to the anomaly, whether they are internal operational changes, seasonal effects, or external events, enabling faster and more informed decision-making.

Amazon Forecast, on the other hand, is primarily a predictive analytics service. It uses historical time-series data to generate accurate forecasts for future metrics, helping organizations anticipate demand, inventory requirements, or trends. While Forecast is very effective at predicting expected values over time, it is not designed to automatically detect anomalies in real-time data. Forecast focuses on trend extrapolation and probabilistic forecasting rather than monitoring for deviations from expected behavior. Therefore, although it provides useful predictive insights, it cannot automatically flag unexpected spikes or drops in metrics without additional custom logic.

SageMaker Autopilot is an automated machine learning (AutoML) service that simplifies the process of building, training, and deploying ML models. It is capable of analyzing datasets, selecting the best algorithms, and producing trained models with minimal human intervention. However, Autopilot does not provide out-of-the-box anomaly detection for ongoing metrics or business KPIs. Any anomaly detection workflow built with Autopilot would require additional configuration, such as designing models specifically for outlier detection, making it less practical for organizations seeking immediate, automated anomaly detection.

AWS Lambda is a serverless compute service that executes code in response to triggers such as data uploads, scheduled events, or API calls. While it is highly flexible and can be integrated into custom monitoring or anomaly detection pipelines, Lambda itself does not offer built-in anomaly detection or machine learning capabilities. Organizations would need to develop and maintain custom logic to detect anomalies, which increases operational overhead and complexity.

Given these comparisons, Lookout for Metrics is the ideal choice for automatic anomaly detection. It is purpose-built for monitoring business metrics, requires minimal configuration, and provides both detection and insight into the causes of anomalies. Its fully managed nature allows organizations to quickly identify and respond to unusual events in their data without the need for custom ML pipelines, making it the most efficient and practical solution among the options listed.

Question 160

A healthcare company wants to label sensitive images securely with HIPAA compliance. Which service is appropriate?

A) SageMaker Ground Truth Private Workforce
B) Mechanical Turk
C) AWS Batch
D) Rekognition Custom Labels

Answer: A

Explanation:

SageMaker Ground Truth Private Workforce is designed to provide a secure and managed environment for human labeling of sensitive data. Unlike public labeling solutions, it allows organizations to create a private workforce that operates entirely within the customer’s VPC, ensuring that sensitive data never leaves a controlled network environment. This architecture is particularly critical for industries such as healthcare, finance, or government, where regulatory requirements demand strict data privacy and security. Ground Truth Private Workforce also integrates seamlessly with other SageMaker services, allowing labeled data to flow directly into ML pipelines for training and model evaluation without additional handling or exposure.

One of the main advantages of using Ground Truth Private Workforce is its compliance capabilities. It supports HIPAA compliance, providing audit logging, controlled access, and monitoring features that are essential when handling medical images or other highly sensitive datasets. The system ensures that only authorized personnel can access the data and that all labeling activities are recorded for audit purposes. This combination of privacy, access control, and compliance makes it a strong choice for organizations that must adhere to strict regulatory requirements.

In comparison, Mechanical Turk is a public crowd-sourcing platform designed for general-purpose labeling tasks. While it is cost-effective and scalable for large volumes of data, it does not provide private workforce isolation or compliance guarantees such as HIPAA. Data submitted to Mechanical Turk may be accessible to a broad pool of workers, which makes it unsuitable for sensitive healthcare or regulated datasets. Mechanical Turk is ideal for low-risk or non-sensitive labeling tasks but is not appropriate when confidentiality and regulatory adherence are required.

AWS Batch is a managed service for running large-scale batch computing workloads. It is highly effective for compute-intensive processing tasks, such as data transformation or large-scale model inference, but it does not provide human labeling capabilities. AWS Batch does not include features for privacy, compliance, or workforce management, so it cannot serve as a solution for annotating sensitive datasets with human oversight.

Rekognition Custom Labels offers automated image labeling using pre-trained or custom ML models. It can significantly reduce human effort by applying machine learning to identify patterns and objects in images. However, it does not involve human annotators and therefore cannot provide the same level of oversight, judgment, or compliance guarantees required for sensitive healthcare data. While suitable for general image classification, Rekognition Custom Labels may not meet regulatory requirements in contexts like medical imaging.

SageMaker Ground Truth Private Workforce is the correct choice because it uniquely combines secure human labeling, integration with ML workflows, and compliance features. It ensures sensitive data is processed safely within a private, managed environment, providing the security, auditability, and privacy controls essential for medical image annotation and other regulated applications.