MLS-C01 Practice Test 31 — 15 Questions

Question 1

A company wants to perform real-time analytics on streaming data from clickstreams. The data needs to be ingested, processed, and made available for querying within seconds. Which AWS service should be used for the processing step?

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

Accepted Answer

Amazon Kinesis Data Analytics. Amazon Kinesis Data Analytics is the correct choice because it enables real-time processing and analysis of streaming data using SQL or Apache Flink. It can ingest data from Kinesis Data Streams or Kinesis Data Firehose, process it with sub-second latency, and output results to destinations like Kinesis Data Streams or Firehose for further querying, meeting the requirement for analytics within seconds.

Answer

AWS Glue

Answer

Amazon Redshift

Answer

Amazon Athena

Question 2

A data scientist is exploring a dataset with 500 features and 10,000 samples. The data scientist computes the pairwise correlation matrix and finds that many features have correlations above 0.9. The data scientist wants to reduce the dataset to 50 features while preserving as much variance as possible. Which technique should be used?

Accepted Answer

Apply Principal Component Analysis (PCA) and keep the top 50 principal components.. Principal Component Analysis (PCA) is the correct technique because it performs an orthogonal linear transformation that projects the original 500 features into a new coordinate system where the axes (principal components) are ordered by the variance they capture. By keeping the top 50 principal components, the data scientist retains the maximum possible variance in the reduced 50-dimensional space, directly addressing the goal of preserving variance while handling high multicollinearity.

Answer

Remove all but one feature from each group of highly correlated features.

Answer

Use Linear Discriminant Analysis (LDA) to project to 50 dimensions.

Answer

Use t-Distributed Stochastic Neighbor Embedding (t-SNE) to reduce to 50 dimensions.

Question 3

Refer to the exhibit. A data engineer has deployed this CloudFormation template. The Glue job 'my-etl-job' reads from the S3 bucket 'my-data-lake-bucket' and writes transformed data to another bucket. After 30 days, the data engineer notices that the Glue job fails with 'Input data not found' errors. What is the most likely cause?

Accepted Answer

The lifecycle configuration deletes objects from the bucket after 30 days, removing the input data.. The lifecycle configuration on the S3 bucket 'my-data-lake-bucket' is set to delete objects after 30 days. Since the Glue job 'my-etl-job' reads input data from this bucket, once the 30-day period elapses, the input data is removed, causing the 'Input data not found' error. This matches the symptom of the job failing after exactly 30 days.

Answer

The temporary directory 'my-temp-dir' is being cleaned up by the lifecycle configuration.

Answer

The script location 's3://my-scripts/etl.py' is being deleted by the lifecycle rule.

Answer

The job bookmark option 'job-bookmark-enable' is causing the job to skip newly arriving data.

Question 4

A data scientist needs to transform raw JSON data from an S3 bucket into Parquet format using AWS Glue. The job must be cost-effective and run only when new data arrives. Which solution should be used?

Accepted Answer

Create a Glue ETL job triggered by an S3 event notification via Lambda.. It uses an S3 event notification to invoke a Lambda function, which then triggers an AWS Glue ETL job only when new data arrives. This event-driven architecture ensures cost-effectiveness by avoiding continuous or scheduled runs, and it directly transforms raw JSON into Parquet format as required.

Answer

Create a Glue crawler that runs continuously.

Answer

Schedule a Glue ETL job to run every hour.

Answer

Use Glue DataBrew to transform data and schedule it daily.

Question 5

A data scientist is analyzing a dataset with missing values in several columns. The dataset contains both numerical and categorical features. Which approach should the data scientist use to handle missing values while minimizing bias and preserving relationships in the data?

Accepted Answer

Use multiple imputation (e.g., MICE) to impute missing values. Multiple imputation (MICE) generates several plausible values for each missing entry by modeling each feature as a function of the others, then pools results across imputations. This preserves the covariance structure and relationships among numerical and categorical variables, and it reflects the uncertainty of the missing data rather than pretending a single value is exact. It is the recommended approach when the goal is minimizing bias in downstream analysis or modeling.

Answer

Use forward-fill to propagate the last observed value

Answer

Delete all rows with missing values

Answer

Replace missing values with the mean or median of each column

Question 6

A team is using SageMaker to train a custom PyTorch model on a large dataset (10 TB) stored in S3. The training job is repeatedly failing due to 'OutOfMemory' errors on the GPU. The team is using a single ml.p3.8xlarge instance. Which change is most likely to resolve the issue?

Accepted Answer

Reduce the batch size in the training script. The 'OutOfMemory' error on the GPU indicates that the model and its associated data exceed the available GPU memory. Reducing the batch size directly decreases the memory footprint per training step, allowing the model to fit within the GPU's memory limits. This is the most direct and effective fix for GPU OOM errors, as it reduces the amount of data processed simultaneously without changing the instance type or input mode.

Answer

Change the instance type to ml.p3.16xlarge (more GPUs)

Answer

Use managed spot training to reduce cost

Answer

Switch the input mode from Pipe to File

Question 7

Which THREE techniques help reduce overfitting in a neural network? (Select THREE.)

Accepted Answer

Dropout. Dropout (A) is correct because randomly deactivating a fraction of neurons during each training step prevents the network from relying on any single neuron's co-adaptation, effectively acting as an ensemble and reducing overfitting. L2 Regularization (B) is correct because adding a penalty proportional to the sum of squared weights (lambda * sum(w^2)) to the loss function constrains weight magnitudes, discouraging the model from fitting noise in the training data. Early Stopping (E) is correct because monitoring validation loss and halting training when it stops improving prevents the model from continuing to memorize the training set and overfit. Increasing the number of layers (C) is not correct because adding depth increases model capacity, which typically worsens overfitting rather than reducing it. Using a larger batch size (D) is not correct because batch size primarily affects gradient estimate variance and training speed, not the model's tendency to overfit in the way these regularization techniques do.

Answer

Increasing the number of layers

Answer

Using a larger batch size

Question 8

A team is using SageMaker to train a deep learning model for image classification. The training job is failing with a 'CUDA out of memory' error. The team is using a p3.2xlarge instance (1 GPU, 16 GB GPU memory). The dataset consists of 256x256 RGB images. Which action is MOST likely to resolve the error without changing the instance type?

Accepted Answer

Reduce the batch size. The 'CUDA out of memory' error indicates that the GPU memory is exhausted. Reducing the batch size directly decreases the memory footprint per training step, allowing the model to fit within the 16 GB GPU memory of the p3.2xlarge instance. This is the most direct and effective fix without changing the instance type.

Answer

Increase the batch size to utilize GPU more efficiently

Answer

Enable automatic model tuning to optimize hyperparameters

Answer

Use Spot Instances to reduce cost

Question 9

A data scientist is training a model using SageMaker and wants to use spot instances to reduce costs. Which THREE considerations should the scientist evaluate? (Choose THREE.)

Accepted Answer

The training job must support checkpointing to save progress.. SageMaker managed spot training requires checkpointing to save model state at regular intervals. If a spot instance is interrupted, the training job can resume from the last checkpoint rather than starting from scratch, which is essential for long-running or expensive training jobs.

Answer

Spot instances have a fixed, lower price than on-demand.

Answer

Spot instances are only available for inference, not training.

Question 10

A data scientist needs to deploy a model with a custom inference container. Which THREE requirements must the container meet for SageMaker hosting?

Accepted Answer

Implement a /ping endpoint for health checks. For SageMaker hosting with a custom inference container, the container must implement the required HTTP contract. Option C is correct because SageMaker sends GET requests to the /ping endpoint to verify the container is healthy and ready to serve traffic. Option D is correct because the container must listen for HTTP requests on port 8080, which is the port SageMaker expects for inference containers. Option E is correct because the /invocations endpoint is where SageMaker sends POST requests containing the input data and expects the model's predictions in response. Option A is incorrect because /opt/ml/input/data is a training data path convention, not a hosting requirement, and no training script is needed for inference. Option B is incorrect because the SageMaker Python SDK is a client-side tool for interacting with SageMaker; the container itself does not need to use it to load the model and can use any framework or library.

Answer

Provide a training script at /opt/ml/input/data

Answer

Use the SageMaker Python SDK to load the model

Question 11

A company uses Amazon Redshift to run analytics on sales data. The data is loaded daily from S3 using COPY commands. The team notices that the COPY command performance degrades over time due to table bloat. The team needs to maintain query performance and reduce storage costs. Which combination of maintenance operations should the team perform regularly? (Choose THREE.)

Accepted Answer

Run the VACUUM command to reclaim space and re-sort rows.. Option C is correct because VACUUM in Amazon Redshift reclaims space from deleted rows and re-sorts rows according to the table's sort keys, directly addressing table bloat and restoring query performance. Option D is correct because a DEEP COPY recreates the table by copying data to a new table with optimal physical storage, fully removing bloat and applying the current sort and distribution settings, which is useful when VACUUM alone is insufficient. Option E is correct because ANALYZE updates table statistics used by the query planner, ensuring efficient query plans after data changes and maintenance operations. Option A is not appropriate because UNLOAD followed by reload is a manual, disruptive workaround rather than a standard maintenance operation for bloat. Option B is not correct because changing the distribution style to KEY is a design decision that does not reclaim space or resolve bloat.

Answer

Run the UNLOAD command to export data to S3 and then reload.

Answer

Change the distribution style of the table to KEY.

Question 12

A company is using Amazon S3 as a data lake. The data engineering team needs to catalog the schema of the data and make it available for querying with Amazon Athena. Which AWS Glue component should be used?

Accepted Answer

AWS Glue Crawlers. AWS Glue Crawlers automatically scan data in Amazon S3, infer the schema, and populate the AWS Glue Data Catalog with metadata tables. This makes the data immediately available for querying with Amazon Athena without manual schema definition.

Answer

AWS Glue Studio

Answer

AWS Glue ETL jobs

Answer

AWS Glue DataBrew

Question 13

A data scientist is using Amazon SageMaker to train a model with a large dataset that does not fit into memory on a single instance. The training algorithm supports distributed training. Which approach should the scientist use to train the model efficiently?

Accepted Answer

Use SageMaker Pipe mode to stream data directly from S3. SageMaker Pipe mode streams data from S3 directly to the training algorithm without writing to disk, enabling processing of large datasets beyond memory.

Answer

Use SageMaker File mode and increase the instance volume size

Answer

Use Amazon EMR to preprocess data and then train on a smaller sample

Answer

Split the data into smaller files and use multiple training jobs sequentially

Question 14

A company is designing a data pipeline to ingest data from multiple sources into an Amazon S3 data lake. The data must be encrypted at rest and in transit. Which TWO actions should be taken to meet these requirements?

Accepted Answer

Enable Server-Side Encryption on the S3 bucket. Enabling Server-Side Encryption (SSE-S3 or SSE-KMS) on the S3 bucket automatically encrypts data at rest when written to disk, using AES-256 encryption. This meets the requirement for encryption at rest without any client-side changes.

Answer

Enable S3 Transfer Acceleration

Answer

Use client-side encryption before uploading

Answer

Use S3 VPC Endpoint

Question 15

A company is using AWS Glue ETL jobs to transform data. The jobs are failing due to insufficient memory. The data processing involves complex joins and aggregations. Which THREE actions can improve job performance and reduce memory usage?

Accepted Answer

Filter and project data early in the transformation to reduce data volume. Option A is correct because filtering rows and projecting only the needed columns early in the Glue ETL script reduces the volume of data read, shuffled, and held in memory during complex joins and aggregations, directly lowering memory pressure. Option C is correct because repartitioning and bucketing organize data by join keys, which minimizes shuffle size and skew, reducing the memory required for shuffle operations in Spark-based Glue jobs. Option D is correct because increasing the number of DPUs (workers) adds more executors and memory capacity, allowing the job to distribute partitions across more nodes and avoid out-of-memory failures on complex transformations. Option B is incorrect because decreasing DPUs reduces available memory and parallelism, worsening the insufficient-memory condition. Option E is incorrect because a single-node cluster concentrates all processing and shuffle on one node, eliminating distributed parallelism and increasing the risk of memory exhaustion.

Answer

Decrease the number of DPUs allocated to the job

Answer

Use a single node cluster to avoid shuffle overhead