AWS ML best practices state that data preprocessing applied during training must be applied identically during inference. For min-max normalization, this requires reusing the minimum and maximum values calculated from the training dataset.
If production data is normalized using different statistics, the feature distributions will differ from what the model learned, leading to degraded prediction accuracy. AWS documentation explicitly warns against recomputing normalization parameters on inference data.
Options A, C, and D introduce data leakage or inconsistent feature scaling. Option B ensures consistency between training and inference pipelines and preserves model integrity.
Therefore, Option B is the correct and AWS-aligned solution.

Amazon SageMaker supports direct file system access for training jobs through Amazon FSx. AWS documentation states that FSx for NetApp ONTAP file systems can be mounted directly to SageMaker training instances when they are located in the same VPC.
Mounting the FSx file system allows SageMaker to stream large datasets efficiently without copying data into Amazon S3. This is ideal for very large datasets such as 6 TB of image data and avoids unnecessary storage duplication.
Options involving SageMaker Data Wrangler are intended for data preparation and exploration, not large- scale training data access. Mountpoint for Amazon S3 is not required and introduces additional complexity.
Therefore, Option A is the correct and AWS-aligned solution.

Amazon SageMaker Pipelines is purpose-built for orchestrating end-to-end ML workflows, including data ingestion, training, evaluation, and deployment. It supports automation, versioning, and deployment of the latest model versions.
MWAA orchestrates general workflows but lacks ML-native features. Lambda cannot handle long-running ML training. Spark processes data but does not manage ML lifecycle.
Therefore, Option A is the correct AWS-native solution.

Option D is correct because Amazon SageMaker AI Random Cut Forest (RCF) is documented by AWS as an unsupervised algorithm for detecting anomalous data points within a data set . The question describes a healthcare use case where the company wants to detect irregularities in patient vital signs and only has an unlabeled dataset . That is a direct match for anomaly detection with an unsupervised method rather than a supervised classification or regression algorithm. AWS documentation also notes that anomalies can appear as unexpected spikes, breaks in patterns, or otherwise unusual observations.
RCF is a strong fit because it does not require labeled target outcomes to begin identifying unusual observations. AWS explains that RCF assigns an anomaly score to each data point, where higher scores indicate potentially anomalous records. This is especially suitable for early warning scenarios in healthcare, where the goal is to surface unusual combinations of vitals, medication history, and lifestyle changes for further review.
The hyperparameter in the answer, num_trees , is a valid RCF hyperparameter and controls the number of trees used in the forest. In AWS examples and documentation, this is one of the key parameters used when configuring the algorithm. By contrast, XGBoost is generally used for supervised tasks, k-means is for clustering rather than anomaly scoring, and DeepAR is for forecasting time series rather than general anomaly detection across an unlabeled multidimensional dataset.
Therefore, the fully verified AWS-docs answer is D , because Random Cut Forest is the AWS-native unsupervised anomaly detection algorithm best aligned to this healthcare irregularity-detection requirement.

Shadow testing allows you to send a copy of live production traffic to a shadow variant of the new model while keeping the existing production model unaffected. This enables you to evaluate the performance of the new model in real-time with live data without impacting end users. SageMaker endpoints support this setup by allowing traffic mirroring to the shadow variant, making it an ideal solution for assessing the new model ' s performance.