CoStudy

HomeCertificationsAWS Certified Machine Learning Engineer MLA-C01 › Deployment and Orchestration of ML Workflows

Deployment and Orchestration of ML Workflows — AWS Certified Machine Learning Engineer MLA-C01 practice questions

50 multiple-choice questions and 10 flashcards on Deployment and Orchestration of ML Workflows, about 24% of the AWS Certified Machine Learning Engineer MLA-C01 bank. Every one carries a written rationale.

Written and maintained by Nick Burton · last updated 2026-08-22 · how we write and review questions

What this chapter covers

Deployment and Orchestration of ML Workflows is one of 4 chapters in CoStudy's AWS Certified Machine Learning Engineer — Associate (MLA-C01) bank, and it holds 50 of the bank's 210 multiple-choice questions — roughly 24% of the total. That proportion is not arbitrary: chapters follow the certifying body's published exam outline, and the number of questions in each is set by that domain's published weight, so the share of your practice time this chapter takes matches the share of the real exam it accounts for.

Studying by chapter is worth doing once you have a diagnostic score. A single overall percentage tells you whether you are close; it does not tell you which domain is dragging. Working a weak chapter in isolation, and re-testing it in isolation, is the fastest way to move a score that has stalled — and it is why the mock exams in CoStudy report by domain rather than as one number.

Free Deployment and Orchestration of ML Workflows practice questions

10 questions drawn from this chapter, with the full rationale shown — the controlling principle behind the right answer, and why each wrong option tempts and fails.

Inference Recommender MOST DIRECTLY provides which value?

  1. Automated data labeling passes
  2. Only endpoint metric charts
  3. Optimal instance-type suggestion
  4. Only IAM policy suggestions

Answer: C — Optimal instance-type suggestion

Inference Recommender load-tests instance types to suggest best cost/performance; the others are different features.

An endpoint receives bursty traffic with long idle gaps between invocations, and cold-start latency of a few seconds is acceptable. Which SageMaker hosting option minimizes idle cost?

  1. Serverless Inference endpoint
  2. Provisioned real-time endpoint sized for peak
  3. SageMaker Batch Transform on a schedule
  4. Multi-Model Endpoint with reserved instances

Answer: A — Serverless Inference endpoint

Serverless Inference scales to zero between invocations, ideal for bursty/idle-heavy traffic; a real-time endpoint sized for peak pays for idle capacity constantly; Batch Transform can't serve live per-request calls; MME with reserved instances still bills for always-on capacity.

For very low-latency inference on custom silicon, MOST fit family is which?

  1. m5 general CPU instances
  2. c5 compute-optimized CPU
  3. inf2 Inferentia instances
  4. p3 general-purpose GPUs

Answer: C — inf2 Inferentia instances

Inferentia is inference-optimized silicon; CPUs are slow; p3 is costly for inference.

Two different frameworks (a preprocessing container and an XGBoost container) must run as one inference call in sequence, with SageMaker managing invocation order. What SageMaker feature is this?

  1. Multi-Model Endpoint (MME)
  2. SageMaker Pipelines ProcessingStep
  3. Production variants for A/B testing
  4. Multi-Container Endpoint (MCE) with an inference pipeline

Answer: D — Multi-Container Endpoint (MCE) with an inference pipeline

MCE (inference pipeline) chains multiple containers so SageMaker passes output from one to the next in a single invocation; MME hosts many independent single-purpose models, not a chained sequence; ProcessingStep is a pipeline build-time step, not live inference chaining; production variants split traffic between competing models, not chain containers.

Limit endpoint invocation to specific principals via which control?

  1. Endpoint policy with IAM
  2. Only S3 bucket policy edits
  3. Only public endpoint URLs
  4. Only Route 53 record edits

Answer: A — Endpoint policy with IAM

Endpoint IAM policies control InvokeEndpoint principals; S3, DNS, and public URLs do not gate access.

SageMaker Clarify PRIMARILY reports which pair of insights?

  1. Bias and explainability results by design
  2. Only IAM policy drift in most cases
  3. Only endpoint idle uptime under this pattern
  4. Only inference cost breakdown for that workload

Answer: A — Bias and explainability results by design

Clarify surfaces bias metrics and SHAP explainability; the others are unrelated to Clarify.

A regulated pipeline must run data validation, training, evaluation, and a human-approval gate as one reproducible, versioned DAG triggered by CI/CD, with lineage tracked automatically. Which service defines the DAG?

  1. AWS Step Functions generic state machine
  2. Amazon Managed Workflows for Apache Airflow (MWAA)
  3. SageMaker Pipelines
  4. AWS Glue Workflows

Answer: C — SageMaker Pipelines

SageMaker Pipelines is purpose-built for ML DAGs with native SageMaker Lineage tracking and Model Registry integration; Step Functions and MWAA can orchestrate ML steps too but lack SageMaker's native lineage/registry integration out of the box; Glue Workflows orchestrates Glue ETL jobs, not full ML training/eval/approval DAGs.

A pipeline's evaluation step must automatically decide whether to proceed to RegisterModelStep based on whether the trained model's AUC exceeds a threshold. Which SageMaker Pipelines construct expresses this branch?

  1. A second, parallel TrainingStep
  2. An EventBridge rule polling S3
  3. A ConditionStep evaluating the metric
  4. A CodePipeline manual approval gate

Answer: C — A ConditionStep evaluating the metric

ConditionStep is the native SageMaker Pipelines construct for branching on a computed metric such as AUC; a parallel TrainingStep doesn't evaluate conditions; polling S3 via EventBridge is an indirect, unnecessary detour for an in-pipeline decision; a CodePipeline manual gate requires a human, not an automatic metric check.

For A/B testing endpoints, MOST fit SageMaker feature is which?

  1. Only Ground Truth jobs
  2. Only Batch Transform runs
  3. Production variants weights
  4. Only multi-model endpoint

Answer: C — Production variants weights

Production variants split traffic between models for A/B; the others do not.

A team wants merges to the main branch of their model-training repo to automatically trigger data validation, training, evaluation, and conditional deployment with no manual pipeline authoring from scratch. Which AWS offering provisions this MLOps scaffold?

  1. SageMaker Projects with the built-in MLOps template
  2. A standalone CodeBuild buildspec only
  3. A manually created Step Functions workflow from scratch
  4. AWS Systems Manager Automation documents

Answer: A — SageMaker Projects with the built-in MLOps template

SageMaker Projects creates the CodeCommit repo, CodePipeline, and CodeBuild wiring in one step via a built-in MLOps template; CodeBuild alone runs commands but has no orchestration; hand-building Step Functions works but isn't the managed scaffold the question asks for; SSM Automation targets infrastructure runbooks, not ML pipeline provisioning.

Deployment and Orchestration of ML Workflows flashcards

4 cards from the 10 in this chapter.

ML workflow CI/CD on AWS — services?

CodeCommit/GitHub + CodeBuild + CodePipeline + SageMaker Pipelines + Model Registry.

SageMaker Inference Recommender?

Benchmarks candidate instance types/configs against a trained model and recommends the best cost/latency fit before production deployment.

SageMaker deployment options?

Real-time endpoint, Asynchronous Inference, Batch Transform, Serverless Inference, Multi-Model Endpoints.

SageMaker Deployment Guardrails?

Built-in traffic-shifting safety net for endpoint updates — canary or linear traffic shift with CloudWatch alarm-based automatic rollback.

Practise the full chapter

These are a sample. The full Deployment and Orchestration of ML Workflows chapter runs 60 items with per-chapter progress tracking, on the web and in the iOS app.

Open AWS Certified Machine Learning Engineer MLA-C01 in CoStudy →

Other AWS Certified Machine Learning Engineer MLA-C01 chapters

All AWS Certified Machine Learning Engineer MLA-C01 practice questions →