Module 1 Quiz: Introduction to MLOps

Test your understanding of MLOps fundamentals with these 30 questions.

---

Question 1

What does MLOps stand for?

1. Machine Learning Online Processing Systems
2. Machine Learning Operations
3. Model Learning Optimization Processes
4. Multiple Learning Operations

---

Question 2

Which of the following best describes the primary goal of MLOps?

1. To replace data scientists with automated systems
2. To train models faster using more GPUs
3. To deploy and maintain ML models in production reliably and efficiently
4. To reduce the cost of cloud computing

---

Question 3

According to industry research, approximately what percentage of ML projects never make it to production?

1. 50%
2. 65%
3. 87%
4. 95%

---

Question 4

Which is NOT a core principle of MLOps?

1. Automation
2. Versioning
3. Manual deployment for better control
4. Continuous monitoring

---

Question 5

At MLOps Maturity Level 0, which of the following is typically true?

1. Automated CI/CD pipelines are in place
2. Models are deployed using Kubernetes
3. Jupyter notebooks are used for everything including production
4. Comprehensive monitoring is implemented

---

Question 6

What is model decay?

1. When model files get corrupted over time
2. When model performance degrades due to changes in data patterns
3. When model weights decrease during training
4. When servers running models become slower

---

Question 7

Which tool is commonly used for data version control in ML projects?

1. Jenkins
2. Prometheus
3. DVC (Data Version Control)
4. Docker

---

Question 8

What is the main difference between DevOps and MLOps?

1. MLOps doesn't use version control
2. MLOps must also version data and models, not just code
3. DevOps is for big companies, MLOps is for startups
4. MLOps doesn't need CI/CD

---

Question 9

Which of the following is a DataOps tool for data quality validation?

1. MLflow
2. Great Expectations
3. Kubernetes
4. TensorFlow

---

Question 10

In the ML lifecycle, what typically comes immediately after feature engineering?

1. Data collection
2. Model deployment
3. Model training
4. Monitoring

---

Question 11

What is a feature store?

1. A retail shop that sells ML software
2. A database for storing model weights
3. A centralized repository for storing and serving features for ML models
4. A type of neural network architecture

---

Question 12

Which MLOps maturity level includes automated CI/CD for ML?

1. Level 0
2. Level 1
3. Level 2
4. Level 3

---

Question 13

What is data drift?

1. When data is stored in multiple locations
2. When the statistical properties of input data change over time
3. When data moves between cloud providers
4. When model predictions become slower

---

Question 14

In a RACI matrix for ML projects, who is typically Responsible for model deployment?

1. Data Scientist
2. Data Engineer
3. MLOps Engineer
4. Product Manager

---

Question 15

What does CI/CD stand for?

1. Continuous Intelligence / Continuous Data
2. Continuous Integration / Continuous Deployment (or Delivery)
3. Code Integration / Code Delivery
4. Centralized Integration / Centralized Deployment

---

Question 16

Which of the following is NOT a common trigger for model retraining?

1. Performance degradation
2. Data drift detection
3. New hire joining the data science team
4. Scheduled time intervals

---

Question 17

What is experiment tracking in MLOps?

1. Tracking which team members are working on experiments
2. Recording parameters, metrics, and artifacts from model training runs
3. Monitoring A/B tests in production
4. Tracking server resource usage

---

Question 18

Which tool is commonly used for experiment tracking?

1. Terraform
2. Apache Kafka
3. MLflow
4. Redis

---

Question 19

What is a model registry?

1. A government database of all ML models
2. A centralized repository for storing, versioning, and managing ML models
3. A type of neural network layer
4. A list of model hyperparameters

---

Question 20

In the context of MLOps, what is "reproducibility"?

1. The ability to copy models from one server to another
2. The ability to print model predictions
3. The ability to recreate the same model results given the same inputs and conditions
4. The ability to run multiple models in parallel

---

Question 21

What is "canary deployment" in the context of ML models?

1. Deploying models that detect anomalies like a canary in a coal mine
2. Gradually routing a small percentage of traffic to a new model version before full rollout
3. Deploying models only in yellow/warning mode
4. Testing models with synthetic data before deployment

---

Question 22

Which of the following is an infrastructure-as-code tool?

1. Pandas
2. Scikit-learn
3. Terraform
4. NumPy

---

Question 23

What is the primary purpose of a data contract between DataOps and MLOps teams?

1. Legal protection for data usage
2. Defining expected schema, quality, and delivery guarantees for data
3. Pricing agreement for cloud storage
4. Employment terms for data engineers

---

Question 24

What does "shadow mode" mean in ML deployment?

1. Running models in dark theme
2. Running a new model alongside production without affecting users, to compare outputs
3. Deploying models without documentation
4. Running models during off-peak hours

---

Question 25

Which metric is commonly used for monitoring model latency?

1. AUC-ROC
2. F1 Score
3. P99 latency (99th percentile response time)
4. Accuracy

---

Question 26

What is the purpose of the problem definition stage in the ML lifecycle?

1. To select the best algorithm
2. To clearly define business objectives, success metrics, and constraints before starting ML work
3. To collect training data
4. To deploy the model

---

Question 27

Which of the following is NOT typically part of model evaluation?

1. Checking AUC-ROC
2. Analyzing confusion matrix
3. Negotiating salary with stakeholders
4. Testing on held-out data

---

Question 28

What is "concept drift"?

1. When the ML algorithm concept becomes outdated
2. When the relationship between input features and the target variable changes over time
3. When team members have different ideas about the project
4. When documentation becomes outdated

---

Question 29

In a centralized MLOps team structure, what is a potential drawback?

1. Too much specialization
2. The team may become a bottleneck and not understand specific team needs
3. Inconsistent standards across teams
4. Too expensive to operate

---

Question 30

What is the recommended approach for team structure in mature MLOps organizations?

1. Only data scientists, no specialized MLOps roles
2. Fully embedded MLOps engineers in each product team
3. Hybrid model with central platform team and ML engineers in product teams
4. Outsource all MLOps to external consultants

---

🎉 Quiz Complete!

You've completed the Module 1 quiz. Review any questions you got wrong and revisit the relevant sections in the module.

Score Interpretation:

• 27-30 correct: Excellent! You have a strong grasp of MLOps fundamentals.
• 22-26 correct: Good understanding, review missed topics.
• 17-21 correct: Fair, recommend re-reading the module.
• Below 17: Review the entire module before proceeding.

---

Next: Module 2: Version Control for ML