MLA-C01 Deployment and Orchestration of ML Workflows Practice Test 4 — 15 Questions

Question 1

An ML platform team deploys a model to a SageMaker real-time endpoint and configures a production variant with a data capture configuration that writes requests and responses to Amazon S3. Weeks later, a data scientist reports that captured files exist but contain only a subset of the fields the model actually received, and some requests appear missing entirely. The endpoint uses a custom inference container that the team built. What is the MOST likely cause?

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

Accepted Answer

The custom container is not logging the full request and response payloads to stdout in the format SageMaker expects, so data capture records only what the container emits.. With bring-your-own containers, SageMaker data capture depends on the container writing request and response data to stdout in the expected JSON format. If the container logs only some fields or skips certain requests, the captured objects in S3 reflect that incomplete output. Endpoint configuration and IAM permissions affect whether capture happens at all, not which fields appear within a captured record.

Answer

Data capture requires the endpoint to use a SageMaker-provided inference container, so a custom container cannot capture complete payloads.

Answer

The endpoint's execution role lacks s3:PutObject permission on the capture bucket, which causes partial writes.

Answer

Data capture only records a sampled percentage of invocations, and the sample rate must be raised to 100 to see every request.

Question 2

A machine learning team wants to deploy the same trained model to multiple SageMaker AI real-time endpoints in different AWS Regions so that each endpoint serves local traffic with low latency. They want to avoid duplicating the model artifact in each Region manually. Which two actions should the engineer take to meet these requirements? (Choose two.)

Accepted Answer

Create the model and endpoint configuration resources independently in each target Region, then create the endpoint in each Region.. SageMaker AI model and endpoint resources are Regional, so serving traffic in multiple Regions requires Regional artifacts and Regional resource definitions. Copying artifacts into each Region and creating the model and endpoint configuration there, then creating the endpoint, produces independent low-latency endpoints.

Answer

Create a single SageMaker AI model resource in the primary Region and reference it from endpoints in every other Region.

Answer

Use SageMaker AI multi-model endpoints in each Region and point them at a shared cross-Region Amazon S3 prefix.

Answer

Enable cross-Region inference on a single endpoint so it automatically routes requests to replicas in other Regions.

Question 3

A retail company runs a nightly retraining workflow built with SageMaker Pipelines. The pipeline includes a processing step that reads from an Amazon S3 input, a training step, and a model registration step. The team reports that the pipeline re-executes the processing step on every run even though the input data and the processing script have not changed, wasting time and cost. They want unchanged steps to be skipped on subsequent executions. (Choose two.)

Accepted Answer

Verify that the processing step's CacheConfig is enabled and that none of the cache-key components, such as the input S3 version or the container image tag, changed between runs.. SageMaker Pipelines caching is driven by CacheConfig and a cache key derived from the step's inputs and configuration. Enabling caching with a key that captures the S3 input and the script, and confirming that none of the key components such as object versions or image tags changed, causes unchanged steps to be skipped. Pipeline name, role, dependencies, and warm pools do not influence cache hits.

Answer

Ensure the pipeline execution uses the same pipeline name and the same role so the cache lookup can find prior outputs.

Answer

Set the processing step's dependency to the training step so the pipeline can infer which outputs are reusable.

Answer

Configure the step to use a warm pool so the container stays running between executions.

Question 4

A data scientist needs to deploy a single ML model that will serve real-time predictions with low latency (under 10 ms) for a high-traffic web application. The model fits in memory and requires GPU acceleration. Which SageMaker inference option is MOST suitable?

Accepted Answer

Real-time endpoint on ml.g4dn instances. Real-time endpoints on GPU instances (ml.g4dn) provide low latency and GPU acceleration, ideal for high-traffic, latency-sensitive workloads.

Answer

Real-time endpoint on ml.m5 instances

Answer

Batch Transform

Answer

Serverless Inference

Question 5

A team has 200 small ML models that need to be served via HTTPS endpoints. Each model is used infrequently, and the team wants to minimize hosting costs. Which SageMaker deployment approach is MOST cost-effective?

Accepted Answer

Use a single multi-model endpoint (MME). A single multi-model endpoint (MME) hosts many models behind one HTTPS endpoint and loads them on demand into shared compute, which is ideal for 200 infrequently used small models because you pay for one endpoint's underlying instances rather than 200 separate ones. SageMaker dynamically loads and unloads models from S3 as requests arrive, dramatically reducing hosting cost while still providing real-time HTTPS inference.

Answer

Use SageMaker Serverless Inference for each model

Answer

Deploy each model on a separate real-time endpoint

Answer

Use Batch Transform for all models

Question 6

An ML team uses SageMaker Pipelines to automate model retraining. They want to skip redundant training steps when input data has not changed. Which feature should they enable?

Accepted Answer

Pipeline caching. SageMaker Pipelines caching stores the output of a step keyed by the step's inputs (code, data, hyperparameters, etc.). When a pipeline run executes and the inputs are unchanged, the cached output is reused and the step is skipped, avoiding redundant training. This directly addresses the requirement to skip training when input data hasn't changed.

Answer

Pipeline variable expressions

Answer

Model registry approval

Answer

Step parallelism

Question 7

A company needs to deploy a new model version to a SageMaker real-time endpoint. They want to route 5% of traffic to the new version initially to monitor for errors before full rollout. Which deployment strategy should they use?

Accepted Answer

Canary deployment with production variants. A canary deployment with production variants allows you to route a specific percentage of traffic (e.g., 5%) to the new model version by adjusting the `InitialVariantWeight` parameter in the production variant configuration. This enables gradual traffic shifting while monitoring errors, and you can later increase the weight to 100% for full rollout. SageMaker real-time endpoints support this natively by hosting multiple model variants behind the same endpoint.

Answer

Blue/green deployment

Answer

Shadow testing

Answer

Multi-model endpoint

Question 8

An ML engineer needs to compile a trained TensorFlow model to run efficiently on a target edge device with an ARM CPU. Which AWS service should they use?

Accepted Answer

SageMaker Neo. SageMaker Neo compiles trained models from frameworks like TensorFlow, PyTorch, and MXNet into optimized executables for specific target hardware, including ARM CPUs, Intel, and AWS Inferentia. It performs graph-level optimizations and generates a runtime that runs efficiently on the edge device.

Answer

SageMaker Debugger

Answer

AWS Inferentia

Answer

Amazon Elastic Inference

Question 9

A data science team uses SageMaker Pipelines for automated training. They need to conditionally register a model only if evaluation metrics exceed a threshold. Which pipeline step type should they use after the evaluation step?

Accepted Answer

Condition step. The Condition step evaluates a condition and branches the pipeline; if the condition is met, the pipeline proceeds to register the model.

Answer

Processing step

Answer

Transform step

Answer

RegisterModel step

Question 10

A company wants to serve a large ensemble of models using NVIDIA Triton Inference Server on SageMaker for high throughput GPU inference. Which SageMaker inference option supports this?

Accepted Answer

Real-time endpoint with a custom container running Triton. NVIDIA Triton Inference Server is a custom inference server that supports multiple frameworks, model ensembles, and dynamic batching. To use it on SageMaker, you deploy a real-time endpoint with a custom container that runs Triton, which supports GPU inference and high throughput. This is the only option that explicitly supports Triton.

Answer

Asynchronous Inference

Answer

Multi-model endpoint

Answer

Serverless Inference

Question 11

A company needs to deploy a model that processes large payloads (up to 1 GB) asynchronously. The results should be written to S3, and the team needs SNS notifications upon completion. Which SageMaker inference option is MOST suitable?

Accepted Answer

Asynchronous Inference. SageMaker Asynchronous Inference is purpose-built for large payloads (up to 1 GB) and long processing times (up to 15 minutes), queuing requests and writing results to S3. It natively supports SNS notifications on completion or failure, matching the requirement exactly. This makes it the correct fit for asynchronous, large-payload processing with S3 output and SNS alerts.

Answer

Batch Transform

Answer

Real-time endpoint

Answer

Serverless Inference

Question 12

An ML team uses AWS Step Functions to orchestrate a retraining pipeline triggered by EventBridge when new training data arrives. The pipeline includes a SageMaker training job and a model evaluation. If evaluation fails, the team wants to send an alert. How should they implement this?

Accepted Answer

Add a Catch rule in the Step Functions state machine to invoke a Lambda alert function. Step Functions supports error handling via Catch rules; a Catch on the training or evaluation task can transition to a Lambda function that sends an alert.

Answer

Use SQS dead-letter queue for failed training jobs

Answer

Configure SageMaker training job to publish to SNS on failure

Answer

Use EventBridge to monitor the training job status

Question 13

A startup wants to deploy a containerized ML application that includes both a model inference server and a preprocessing component in the same endpoint. Which SageMaker endpoint type supports running multiple containers?

Accepted Answer

Multi-container endpoint. SageMaker multi-container endpoints allow you to run up to 15 containers on a single endpoint, with containers invoked in a defined sequence (inference pipeline) or directly. This is the correct choice when a preprocessing component and a model inference server must coexist in one endpoint, since the preprocessing container can transform the request before passing it to the inference container. This pattern is ideal for encapsulating feature engineering with the model.

Answer

Asynchronous Inference

Answer

Multi-model endpoint

Answer

Real-time endpoint

Question 14

A company uses SageMaker to deploy a model and wants to perform A/B testing by splitting traffic between two model variants. Which TWO actions should they take? (Select TWO.)

Accepted Answer

Configure two production variants on the endpoint, each with an initial weight. Option A is correct because SageMaker A/B testing is implemented by configuring a single endpoint with two production variants, each assigned an InitialVariantWeight so the endpoint can split inference traffic between them. Option C is correct because after analyzing the variants' performance, you can shift traffic by calling UpdateEndpointWeightsAndCapacity to change each variant's VariantWeight and DesiredInstanceCount without redeploying the endpoint. Option B is not required: Model Registry approval is a governance step and does not itself create or split endpoint traffic. Option D is wrong because deploying separate endpoints with Route53 weighted routing is a DNS-level workaround, not SageMaker's native A/B testing mechanism. Option E is wrong because shadow testing mirrors live traffic to a new variant for evaluation without serving its responses, so it does not split production traffic between two variants.

Answer

Use SageMaker Model Registry to approve both variants

Answer

Deploy each variant to a separate endpoint and use Route53 weighted routing

Answer

Enable shadow testing on the endpoint

Question 15

An organization wants to automate ML retraining using an event-driven architecture. Which THREE services should they combine? (Select THREE.)

Accepted Answer

SageMaker (training jobs or pipelines). Amazon EventBridge [CORRECT] is the event-driven backbone: it receives events (e.g., new data in S3, a schedule, or a custom PutEvents call) and routes them to targets via rules, which is exactly what an event-driven retraining architecture requires. AWS Lambda [CORRECT] acts as the lightweight, serverless compute glue that responds to those EventBridge events and invokes the training workflow, so no servers need to be managed. SageMaker (training jobs or pipelines) [CORRECT] performs the actual model retraining and orchestration, since SageMaker Pipelines can be triggered programmatically to run training jobs and register updated models. AWS Glue is a data-integration/ETL service and, while useful for preparing data, it is not required to deliver the event-driven retraining trigger or the training itself. Amazon CloudWatch Logs is for storing and monitoring log data and does not provide event routing or ML training capability.

Answer

AWS Glue

Answer

Amazon CloudWatch Logs