MLS-C01 Mock Exam 72 — 25 Questions

Question 1

A company stores sensitive customer data in Amazon S3 and uses AWS Glue to catalog and transform the data. The data must be encrypted at rest with a customer-managed AWS KMS key to meet compliance requirements. A data engineer creates an AWS Glue crawler and an ETL job. What should the engineer do to ensure the Glue ETL job can read and write the encrypted data?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Grant the Glue ETL job's IAM role permissions to use the KMS key for encrypt and decrypt operations.. When S3 objects are encrypted with a customer-managed KMS key, any service that reads or writes those objects must have permissions to use the key. For AWS Glue ETL jobs, the IAM role associated with the job must include kms:Decrypt and kms:Encrypt permissions on the specific KMS key. This ensures the job can access the data while maintaining compliance.

Answer

Configure the Glue crawler to use a custom classifier that handles encrypted data.

Answer

Attach a bucket policy to the S3 bucket that allows the Glue service principal to decrypt objects.

Answer

Enable server-side encryption with Amazon S3-managed keys (SSE-S3) on the S3 bucket.

Question 2

A data engineer must give a SageMaker training job temporary, least-privilege access to a specific Amazon S3 prefix containing training data, without storing long-term credentials anywhere. The training job runs on a SageMaker-managed instance and the engineer controls only the job configuration. What is the correct approach?

Accepted Answer

Create an IAM role with an S3 access policy scoped to the prefix and specify its ARN as the execution role when creating the training job.. SageMaker training jobs authenticate using an IAM execution role. When the job starts, SageMaker assumes that role and delivers rotating temporary credentials to the container, so nothing long-lived is stored. Pointing the role's policy at the specific S3 prefix restricts access to exactly the required data, satisfying least privilege while needing only the role ARN in the job definition.

Answer

Attach an S3 bucket policy that allows the SageMaker service principal to read the prefix and rely on that alone.

Answer

Generate an IAM user access key and pass it to the training container through hyperparameters.

Answer

Store the credentials in AWS Secrets Manager and have the training script fetch them at startup.

Question 3

A machine learning engineer is deploying a real-time fraud detection model on Amazon SageMaker. The model is a PyTorch neural network that requires GPU acceleration. The engineer needs to ensure that the endpoint automatically scales the number of instances based on the number of invocations per instance. The engineer has already created the SageMaker model and endpoint configuration. Which scaling policy should the engineer configure to meet these requirements?

Accepted Answer

Target tracking scaling policy with a predefined metric of SageMakerVariantInvocationsPerInstance.. The correct answer is the target tracking scaling policy using the predefined SageMakerVariantInvocationsPerInstance metric because it directly scales based on the number of invocations per instance, which is exactly what the engineer needs. This policy is simple to configure and automatically adjusts capacity to maintain the target value, ensuring optimal performance and cost efficiency for the real-time fraud detection endpoint.

Answer

Step scaling policy with a CloudWatch alarm on the CPUUtilization metric of the endpoint instances.

Answer

Target tracking scaling policy with a custom metric published to Amazon CloudWatch that measures the number of invocations per instance.

Answer

Scheduled scaling policy that increases the instance count during peak business hours.

Question 4

A financial institution is building a fraud detection model using Amazon SageMaker. The dataset is highly imbalanced, with only 0.5% of transactions being fraudulent. The team wants to maximize the detection of fraudulent transactions while keeping false positives at an acceptable level. They decide to use the SageMaker built-in XGBoost algorithm. Which evaluation metric and corresponding hyperparameter tuning strategy should they use to achieve their goal?

Accepted Answer

Optimize for F1 score and tune scale_pos_weight along with max_depth.. In fraud detection with extreme imbalance, the F1 score provides a balance between precision and recall, which is crucial for maximizing fraud detection while keeping false positives acceptable. XGBoost's scale_pos_weight hyperparameter adjusts the loss function to give more weight to the positive class, effectively addressing imbalance. Tuning max_depth controls model complexity, preventing overfitting to the minority class. This strategy is widely recommended for imbalanced binary classification.

Answer

Optimize for precision and tune subsample and colsample_bytree.

Answer

Optimize for accuracy and tune the number of rounds and learning rate.

Answer

Optimize for area under the ROC curve (AUC) and tune max_depth and eta.

Question 5

A machine learning team is using Amazon SageMaker to train a custom model with a deep neural network. They notice that the training loss decreases steadily, but the validation loss starts to increase after a few epochs. The team wants to improve the model's generalization performance without changing the model architecture. Which technique should they apply?

Accepted Answer

Implement early stopping based on validation loss.. The scenario describes classic overfitting: training loss decreases while validation loss increases. Early stopping is a regularization technique that halts training at the point of best validation performance, preventing the model from memorizing noise in the training data. It requires no architectural changes and is straightforward to implement, making it the ideal choice.

Answer

Decrease the batch size.

Answer

Increase the learning rate.

Answer

Add dropout layers to the neural network.

Question 6

A company stores raw sensor readings in Amazon S3 and must produce a cleaned, feature-engineered dataset for model training. The transformations are written in PySpark, the job must run on a schedule without managing servers, and the output must be immediately queryable by Amazon Athena. Which service should the data engineer use?

Accepted Answer

AWS Glue with a Spark ETL job that writes the transformed output to S3 in Parquet format.. AWS Glue provides managed, serverless Spark execution with job scheduling and a native path to the Data Catalog, so PySpark transformations can run on a schedule and land Parquet in S3 that Athena queries immediately. Lambda, EMR on EC2, and streaming Flink each violate at least one requirement around scale, server management, or batch semantics.

Answer

Amazon Kinesis Data Analytics with a Flink application writing results to S3.

Answer

Amazon EMR on Amazon EC2 with a long-running cluster and a nightly cron job.

Answer

AWS Lambda with a Python script that processes each S3 object on upload.

Question 7

A data scientist is using Amazon SageMaker to train a linear regression model on a dataset with a single feature. The model's training loss is high, and the validation loss is similarly high, indicating underfitting. The data scientist suspects that the relationship between the feature and the target is not linear. Which action should the data scientist take to improve the model's performance?

Accepted Answer

Add polynomial features (e.g., squared and cubic terms) to the input data and retrain the linear model.. Underfitting with a linear model on a single feature suggests the true relationship is nonlinear. Adding polynomial features increases the hypothesis space, enabling the linear model to fit curves. This directly addresses the high bias. Other options either target overfitting or do not increase model capacity. Thus, polynomial feature engineering is the most effective solution.

Answer

Apply L2 regularization (ridge) to the linear regression model to reduce overfitting.

Answer

Collect more training data to provide more examples for the model to learn from.

Answer

Increase the learning rate for the optimization algorithm to converge faster.

Question 8

A data scientist is performing exploratory data analysis on a dataset stored in Amazon S3. The dataset contains sensitive customer information and must remain encrypted at rest with a customer-managed AWS KMS key. The scientist needs to run a Jupyter notebook in Amazon SageMaker Studio to compute summary statistics and generate visualizations without downloading the data. Which approach allows the scientist to access the data securely while meeting the encryption requirement?

Accepted Answer

Attach an IAM role to the SageMaker Studio notebook instance that grants s3:GetObject and kms:Decrypt permissions for the specific S3 bucket and KMS key, then read the data directly in the notebook using the AWS SDK for Python (Boto3).. The correct approach is to attach an IAM role to the SageMaker Studio notebook instance with permissions for s3:GetObject and kms:Decrypt, then use Boto3 to read the data directly. This ensures the data remains encrypted at rest and is decrypted only in memory during analysis, meeting the security requirement. Other options either fail to handle KMS permissions correctly or introduce unnecessary steps.

Answer

Use Amazon Athena from within the SageMaker Studio notebook to query the S3 data. Athena automatically handles KMS decryption and returns results as a pandas DataFrame without additional permissions.

Answer

Copy the dataset to an Amazon EBS volume attached to the SageMaker Studio notebook instance, then decrypt it using the KMS key. The EBS volume inherits the same encryption as the S3 bucket.

Answer

Use SageMaker Studio's built-in integration with AWS Glue DataBrew to read the S3 data directly, because DataBrew automatically decrypts KMS-encrypted objects.

Question 9

While performing EDA on a dataset of 2 million customer records, a data scientist computes a correlation matrix and finds two features with a Pearson coefficient of 0.97. Both features are used in a linear regression model, and the model's coefficients are large and flip sign between retraining runs. Which action should the data scientist take?

Accepted Answer

Recognize the multicollinearity and drop one of the two features, or combine them into a single feature such as their average or principal component.. A 0.97 correlation between two predictors indicates near-redundancy, which inflates the variance of the estimated coefficients and produces the large, sign-flipping values observed across retraining runs. Dropping one feature or combining the pair into a single derived variable removes the redundancy and yields stable, interpretable coefficients, addressing the root cause rather than the symptom.

Answer

Standardize both features to zero mean and unit variance, which removes the correlation between them and stabilizes the coefficients.

Answer

Add more training records until the correlation coefficient drops below 0.5, then refit the model with both features retained.

Answer

Keep both features and increase the regularization strength of the linear model to zero out the unstable coefficients.

Question 10

A data scientist needs to store a large dataset in Amazon S3 for a machine learning project. The dataset will be accessed frequently by multiple Amazon SageMaker training jobs and notebooks. The data must be highly durable and available, and the solution should be cost-effective for frequent access. Which S3 storage class should the data scientist use?

Accepted Answer

S3 Standard. The dataset is accessed frequently by SageMaker training jobs and notebooks, so the storage class must provide low latency, high durability, and no retrieval fees. S3 Standard meets these requirements and is cost-effective for frequent access despite a higher storage cost per GB. Other classes either charge retrieval fees or have slow retrieval, making them unsuitable.

Answer

S3 One Zone-Infrequent Access (S3 One Zone-IA)

Answer

S3 Standard-Infrequent Access (S3 Standard-IA)

Answer

S3 Glacier Deep Archive

Question 11

A company is deploying a real-time inference endpoint on Amazon SageMaker. They need to ensure the endpoint can automatically scale based on incoming traffic and that the scaling policy targets a specific utilization metric. Which two actions should they take? (Choose two.)

Accepted Answer

Configure an Application Auto Scaling policy for the SageMaker endpoint variant.. To automatically scale a SageMaker endpoint, you must configure Application Auto Scaling and define a scaling policy. Target tracking with the SageMakerVariantInvocationsPerInstance metric is a common and effective policy that adjusts instance count to maintain a target invocation rate. Together, these actions provide dynamic scaling without manual intervention, meeting the company's need.

Answer

Enable Elastic Inference on the endpoint to handle traffic spikes.

Answer

Use the SageMaker UpdateEndpoint API to manually adjust the instance count.

Answer

Create a CloudWatch alarm that triggers an AWS Lambda function to call UpdateEndpoint.

Question 12

An ML engineer trains a model on Amazon SageMaker using a dataset stored in Amazon S3. The training job completes successfully, but the engineer notices that the model artifacts are not saved to the specified S3 output path. The training job's IAM role has full access to the input S3 bucket but only read access to the output bucket. What is the most likely cause of the missing model artifacts?

Accepted Answer

The training job's IAM role lacks s3:PutObject permission for the output S3 bucket.. SageMaker training jobs need both read access to input data and write access to the output S3 bucket. The role's read-only permission on the output bucket prevents the job from uploading model artifacts. Granting s3:PutObject on the output prefix ensures artifacts are stored. The other options either misstate the permissions or incorrectly attribute the cause to missing configuration or local filesystem issues.

Answer

The training job's IAM role lacks s3:GetObject permission for the input S3 bucket.

Answer

The training container does not have write access to the local filesystem to save the model.

Answer

The SageMaker training job did not specify an output S3 path in the estimator configuration.

Question 13

A company runs a nightly Apache Spark job on an Amazon EMR cluster to process log files stored in Amazon S3. The job writes its output back to S3 and then the cluster is terminated. The data engineering team wants the job to read and write S3 data using the same API as a Hadoop file system, with the connector maintained by AWS and optimized for the EMR File System. Which storage connector should the team use?

Accepted Answer

s3:// with EMRFS (EMR File System). EMRFS is the AWS-maintained file system implementation that lets EMR clusters read and write Amazon S3 through the s3:// scheme using Hadoop FileSystem APIs. It is optimized for EMR, supports consistent view for list-after-write consistency, and is the standard connector for Spark jobs that persist output to S3 before the cluster terminates.

Answer

s3n:// with the legacy S3 Native File System

Answer

s3a:// with the open-source S3A connector

Answer

hdfs:// pointing at the cluster's local HDFS

Question 14

A data scientist wants to quickly explore a 50 MB CSV file of customer attributes inside an Amazon SageMaker Studio notebook. They need summary statistics, column data types, and a correlation view with minimal code, and the data fits comfortably in memory. Which approach is most appropriate?

Accepted Answer

Load the CSV into a pandas DataFrame with read_csv and call the describe method, then render a correlation heatmap in the notebook.. When the dataset is small enough to fit in notebook memory, loading it with pandas and using describe plus a correlation heatmap delivers summary statistics, inferred data types, and pairwise relationships with almost no setup. Distributed or catalog-based services add latency and cost that are unjustified at 50 MB, so the direct in-memory approach is the pragmatic choice.

Answer

Upload the CSV to Amazon S3, crawl it with an AWS Glue crawler, and query it with Amazon Athena to get column statistics.

Answer

Convert the CSV to TFRecord format and use TensorFlow Data Validation to generate statistics before any modeling.

Answer

Provision an Amazon Redshift cluster, COPY the CSV into a table, and run SQL aggregate queries to obtain the summary statistics.

Question 15

A data scientist is analyzing a binary classification dataset and observes that one feature has a highly skewed distribution with several extreme values. The scientist wants a visualization that shows how the central tendency and spread of this feature differ between the two classes without being dominated by the extreme observations. Which visualization is most appropriate?

Accepted Answer

A pair of box plots of the feature, one for each class.. The requirement is a per-class comparison of central tendency and spread that resists distortion from extreme values. Box plots summarize each class using the median and interquartile range, which are not pulled by outliers, and placing the two boxes side by side makes the differences easy to read. Alternative charts either lose the per-class distributional detail or are dominated by the skewed tail.

Answer

A pie chart of the class proportions in the dataset.

Answer

A scatter plot of the feature against the row index.

Answer

A single histogram of the feature with all classes overlaid.

Question 16

A financial institution is training a binary classification model to detect fraudulent transactions. The dataset is highly imbalanced, with only 0.5% of transactions being fraudulent. The data scientist uses Amazon SageMaker's built-in XGBoost algorithm with default hyperparameters and achieves 99.5% accuracy on the validation set, but the model fails to identify any fraudulent cases. Which modification to the training process is MOST likely to improve the model's ability to detect fraud?

Accepted Answer

Use the scale_pos_weight hyperparameter in XGBoost to assign a higher weight to the positive (fraudulent) class.. The model achieves high accuracy by predicting all transactions as non-fraudulent, which is useless. The core problem is the severe class imbalance, causing the loss function to favor the majority class. Using scale_pos_weight in XGBoost increases the weight of the positive class, forcing the model to learn from fraudulent examples. This directly improves the model's ability to detect fraud without changing the algorithm or data pipeline.

Answer

Deploy the model to a real-time endpoint and use a higher threshold for classification.

Answer

Increase the number of training epochs to allow the model to learn more from the minority class.

Answer

Apply principal component analysis (PCA) to reduce dimensionality before training.

Question 17

A machine learning platform team stores training data in Amazon S3 and uses AWS Lake Formation to manage access. Data scientists in one business unit must read only the columns relevant to their models from a shared table, while other columns containing personally identifiable information must remain hidden. The team also wants to enforce these restrictions consistently across Athena and Amazon Redshift Spectrum. What should the team do?

Accepted Answer

Configure Lake Formation column-level permissions and grant SELECT on specific columns to the data scientist role.. Lake Formation column-level permissions are the native mechanism to grant SELECT on a subset of columns while hiding PII, and they are enforced across integrated engines including Athena and Redshift Spectrum. IAM, presigned URLs, and table splitting cannot express column-level access within a shared table.

Answer

Enable S3 Block Public Access and require data scientists to use presigned URLs.

Answer

Create an IAM policy that denies access to the S3 prefixes containing the PII columns.

Answer

Use AWS Glue crawlers to create separate tables for each column and grant access per table.

Question 18

A data engineer must give a SageMaker training job read access to a specific S3 bucket containing training data, while ensuring the job cannot access any other bucket. The training job runs with an IAM role. What is the MOST secure way to grant this access?

Accepted Answer

Attach an IAM policy to the training job's execution role that allows s3:GetObject and s3:ListBucket on the specific bucket and its objects.. Scoping an IAM policy to the specific bucket and attaching it to the training job's execution role grants only the required read permissions. SageMaker assumes the role to obtain temporary credentials, so no long-lived keys are needed, and the policy prevents access to other buckets. Broad managed policies, embedded keys, or public buckets all violate the least-privilege and security requirements.

Answer

Attach the AmazonS3ReadOnlyAccess managed policy to the training job's execution role so it can read any bucket in the account.

Answer

Make the S3 bucket public and restrict access using an S3 bucket policy that allows only the training job's IP address range.

Answer

Store long-term AWS access keys in the training container's environment variables and use them to access the bucket.

Question 19

A media company stores millions of small thumbnail images in an Amazon S3 bucket and wants to run a nightly machine learning preprocessing job that reads all objects, resizes them, and writes results to another bucket. The team observes that the job is slow because each small object read incurs request overhead, and they want to reduce the number of GET requests. Which approach should the data engineer recommend?

Accepted Answer

Aggregate the small objects into larger files using a format such as Apache Parquet or TFRecord, then have the preprocessing job read the consolidated files.. Reading millions of tiny objects from Amazon S3 incurs one GET request per object, and that request overhead dominates runtime. Consolidating the small files into larger objects, ideally in a columnar or record-oriented format, reduces the request count and increases bytes per request, which is the standard remedy for the small-file problem in data lakes.

Answer

Use Amazon S3 Inventory to list objects and then read each object individually with the S3 Select API.

Answer

Increase the S3 request rate by adding a random prefix to each object key and re-uploading the objects.

Answer

Enable S3 Transfer Acceleration on the source bucket to speed up each small object read.

Question 20

During EDA a data scientist computes summary statistics and notices that a feature has a mean of 52 and a median of 12, with a long tail of large values. Before deciding on preprocessing, the scientist wants a single robust, scale-independent measure of how asymmetric the distribution is. Which statistic should be computed?

Accepted Answer

The quartile skewness coefficient, (Q3 + Q1 - 2*Q2) / (Q3 - Q1).. The quartile skewness coefficient depends only on the three quartiles, making it resistant to the long tail of large values while still being scale-independent and signed to indicate direction. It directly quantifies the asymmetry that the mean-median gap suggests, giving a robust single-number summary before preprocessing decisions are made.

Answer

The interquartile range of the feature.

Answer

The Pearson moment coefficient of skewness using the third standardized moment.

Answer

The median absolute deviation of the feature.