WGU D501 MACHINE LEARNING DEVOPS
PROJECT 1 AND 2 BUNDLE [QUESTIONS
1-100] AND ANSWERS UPDATED
2026/2027 | DETAILED RATIONALES
Study-use notice: This is an original practice question
bank for studying D501 concepts. It is not an official
WGU assessment, answer key, or guaranteed submission
solution. Exact project requirements should be checked
against your current WGU course materials and rubric.
INTRODUCTION
WGU D501 Machine Learning DevOps focuses on applying machine-learning and
software-engineering principles within a reproducible operational workflow. The
course is particularly relevant to learners who need to understand how a machine-
learning solution moves beyond experimentation toward reliable development, testing,
deployment, monitoring, and maintenance. Rather than treating a model as an isolated
algorithm, Machine Learning DevOps emphasizes the broader lifecycle surrounding
data, code, model artifacts, environments, automation, version control, and
operational performance. Projects associated with this subject require students to
reason about practical implementation decisions and demonstrate that an ML solution
can be developed and managed systematically. This practice bank is designed to
reinforce those skills through difficult, scenario-based questions. The questions
emphasize lifecycle design, reproducibility, source control, testing, CI/CD, data and
model versioning, containerization, deployment, monitoring, drift, experiment
tracking, and troubleshooting. Each question includes a detailed rationale explaining
both the correct choice and why the alternatives are less appropriate. Use the
questions to identify conceptual gaps, then verify project-specific requirements
against the current WGU instructions before submitting academic work.
CORE DOMAINS TESTED
1.
Machine Learning Lifecycle — Planning, developing, validating, deploying,
and maintaining ML solutions.
2.
3.
MLOps Principles — Reproducibility, automation, collaboration, governance,
and operational reliability.
1
, 4.
5.
Version Control — Managing source code, configurations, datasets, and
model artifacts.
6.
7.
Data Engineering for ML — Data validation, preprocessing, pipelines,
lineage, and reproducibility.
8.
9.
Model Development — Training, validation, evaluation, feature engineering,
and experiment management.
10.
11.
CI/CD for ML — Automated testing, builds, integration, deployment, and
release controls.
12.
13.
Containers and Environments — Dependency isolation, reproducible
environments, and deployment consistency.
14.
15.
Model Deployment — Batch and real-time inference, APIs, serving
architectures, and rollback.
16.
17.
Monitoring — Model performance, data quality, latency, resource utilization,
and operational health.
18.
19.
Model/Data Drift — Detecting distribution changes and determining when
retraining is warranted.
20.
2
, 21.
Security and Governance — Secrets management, access control,
auditability, and responsible ML operations.
22.
23.
Troubleshooting and Reliability — Diagnosing failures across data, code,
infrastructure, and models.
24.
QUESTIONS 1-100
Q1: A machine-learning team obtains excellent validation performance locally, but
another developer cannot reproduce the model because the project depends on
unrecorded Python package versions and manually configured environment variables.
Which MLOps practice would most directly address the problem?
A) Increase the number of model features
B) Define and version the environment and configuration dependencies
C) Increase the training dataset size
D) Replace the validation dataset with the production dataset
Rationale: The correct answer is B because reproducibility requires the software
environment, dependencies, configuration, and other relevant conditions to be
explicitly defined and versioned. Option A may improve predictive performance but
does not solve environmental inconsistency. Option C addresses statistical learning
rather than reproducibility. Option D would introduce evaluation contamination and
is not an appropriate solution.
Q2: A team stores its application source code in Git but keeps trained models only on
individual developers' laptops. A production deployment fails because the developer
who created the model is unavailable. What is the most appropriate improvement?
A) Retrain the model manually whenever deployment fails
B) Store model artifacts in a controlled, versioned artifact repository
C) Put the model binary directly into every source-code commit
D) Remove model versioning because the training script can recreate everything
Rationale: The correct answer is B because trained models are deployment artifacts
and should be managed in a controlled repository with identifiable versions. Option A
creates unnecessary manual dependency. Option C can make source repositories
unnecessarily large and difficult to manage. Option D assumes training is perfectly
deterministic and ignores the need to preserve the exact artifact that was deployed.
3
PROJECT 1 AND 2 BUNDLE [QUESTIONS
1-100] AND ANSWERS UPDATED
2026/2027 | DETAILED RATIONALES
Study-use notice: This is an original practice question
bank for studying D501 concepts. It is not an official
WGU assessment, answer key, or guaranteed submission
solution. Exact project requirements should be checked
against your current WGU course materials and rubric.
INTRODUCTION
WGU D501 Machine Learning DevOps focuses on applying machine-learning and
software-engineering principles within a reproducible operational workflow. The
course is particularly relevant to learners who need to understand how a machine-
learning solution moves beyond experimentation toward reliable development, testing,
deployment, monitoring, and maintenance. Rather than treating a model as an isolated
algorithm, Machine Learning DevOps emphasizes the broader lifecycle surrounding
data, code, model artifacts, environments, automation, version control, and
operational performance. Projects associated with this subject require students to
reason about practical implementation decisions and demonstrate that an ML solution
can be developed and managed systematically. This practice bank is designed to
reinforce those skills through difficult, scenario-based questions. The questions
emphasize lifecycle design, reproducibility, source control, testing, CI/CD, data and
model versioning, containerization, deployment, monitoring, drift, experiment
tracking, and troubleshooting. Each question includes a detailed rationale explaining
both the correct choice and why the alternatives are less appropriate. Use the
questions to identify conceptual gaps, then verify project-specific requirements
against the current WGU instructions before submitting academic work.
CORE DOMAINS TESTED
1.
Machine Learning Lifecycle — Planning, developing, validating, deploying,
and maintaining ML solutions.
2.
3.
MLOps Principles — Reproducibility, automation, collaboration, governance,
and operational reliability.
1
, 4.
5.
Version Control — Managing source code, configurations, datasets, and
model artifacts.
6.
7.
Data Engineering for ML — Data validation, preprocessing, pipelines,
lineage, and reproducibility.
8.
9.
Model Development — Training, validation, evaluation, feature engineering,
and experiment management.
10.
11.
CI/CD for ML — Automated testing, builds, integration, deployment, and
release controls.
12.
13.
Containers and Environments — Dependency isolation, reproducible
environments, and deployment consistency.
14.
15.
Model Deployment — Batch and real-time inference, APIs, serving
architectures, and rollback.
16.
17.
Monitoring — Model performance, data quality, latency, resource utilization,
and operational health.
18.
19.
Model/Data Drift — Detecting distribution changes and determining when
retraining is warranted.
20.
2
, 21.
Security and Governance — Secrets management, access control,
auditability, and responsible ML operations.
22.
23.
Troubleshooting and Reliability — Diagnosing failures across data, code,
infrastructure, and models.
24.
QUESTIONS 1-100
Q1: A machine-learning team obtains excellent validation performance locally, but
another developer cannot reproduce the model because the project depends on
unrecorded Python package versions and manually configured environment variables.
Which MLOps practice would most directly address the problem?
A) Increase the number of model features
B) Define and version the environment and configuration dependencies
C) Increase the training dataset size
D) Replace the validation dataset with the production dataset
Rationale: The correct answer is B because reproducibility requires the software
environment, dependencies, configuration, and other relevant conditions to be
explicitly defined and versioned. Option A may improve predictive performance but
does not solve environmental inconsistency. Option C addresses statistical learning
rather than reproducibility. Option D would introduce evaluation contamination and
is not an appropriate solution.
Q2: A team stores its application source code in Git but keeps trained models only on
individual developers' laptops. A production deployment fails because the developer
who created the model is unavailable. What is the most appropriate improvement?
A) Retrain the model manually whenever deployment fails
B) Store model artifacts in a controlled, versioned artifact repository
C) Put the model binary directly into every source-code commit
D) Remove model versioning because the training script can recreate everything
Rationale: The correct answer is B because trained models are deployment artifacts
and should be managed in a controlled repository with identifiable versions. Option A
creates unnecessary manual dependency. Option C can make source repositories
unnecessarily large and difficult to manage. Option D assumes training is perfectly
deterministic and ignores the need to preserve the exact artifact that was deployed.
3
