Databricks Certification Practice Test 10

Practice Test 10

Question 1 of 30

📋 Multiple Choice

Question 201

## Question 201 A retail company has trained a demand forecasting model. The model performs well in development but business users cannot reproduce previous prediction results after a retraining cycle. What is the BEST action?

A. Create larger clusters

✓

B. Track model versions

C. Increase partitions

D. Run OPTIMIZE

Answer

✅ B. Track model versions

Explanation

Model versioning ensures engineers can identify, reproduce and compare model outputs over time. ---

Answered: 0 / 30

Databricks Certification Practice Test 10

Practice Test 10

Insightful Saga — Modern Data Engineering Certification Preparation

Question 201

A. Create larger clusters

B. Track model versions

C. Increase partitions

D. Run OPTIMIZE

Answer: ✅ B. Track model versions

Explanation:

Model versioning ensures engineers can identify, reproduce and compare model outputs over time. ---

Question 202

## Question 202 A bank has multiple data scientists training fraud detection models. Management wants all experiments tracked in a central location. What is the BEST solution?

A. Store results in Excel

B. Email experiment metrics

C. Use MLflow Experiment Tracking

D. Create more notebooks

Answer: ✅ C. Use MLflow Experiment Tracking ---

Question 203

## Question 203 A telecom company has 500 machine learning features stored in different notebooks. Data scientists frequently calculate the same features multiple times. What is the BEST action?

A. Build a Feature Store

B. Add more clusters

C. Create additional workspaces

D. Store features in CSV files

Answer: ✅ A. Build a Feature Store ---

Question 204

## Question 204 A recommendation model suddenly becomes less accurate. Customer behavior changed significantly during the last month. What is the MOST likely issue?

A. Schema Drift

B. Data Drift

C. Photon Failure

D. Notebook Timeout

Answer: ✅ B. Data Drift ---

Question 205

## Question 205 A healthcare company must explain why an AI model approved one patient while rejecting another. What is the BEST approach?

A. Remove model documentation

B. Improve model explainability

C. Delete predictions

D. Hide training logic

Answer: ✅ B. Improve model explainability ---

Question 206

## Question 206 A data science team repeatedly trains models using different hyperparameter values. What is the BEST capability to track?

A. Cluster Names

B. Experiment Parameters

C. Dashboard Colors

D. Catalog Structure

Answer: ✅ B. Experiment Parameters ---

Question 207

## Question 207 A manufacturing company wants machine learning models automatically retrained when performance drops below a threshold. What is the BEST approach?

A. MLOps Pipeline

B. Manual Emails

C. CSV Exports

D. More SQL Warehouses

Answer: ✅ A. MLOps Pipeline ---

Question 208

## Question 208 A credit scoring model predicts significantly better in training than production. What should engineers investigate first?

A. Overfitting

B. Notebook Font

C. Cluster Policy

D. Unity Catalog

Answer: ✅ A. Overfitting ---

Question 209

## Question 209 A company wants consistent customer segmentation features shared across dozens of ML models. What is the BEST strategy?

A. Duplicate Logic

B. Feature Store

C. Manual Feature Creation

D. Separate Definitions

Answer: ✅ B. Feature Store ---

Question 210

## Question 210 A machine learning pipeline fails because expected columns are missing from incoming datasets. What should be implemented?

A. Feature Validation

B. Dashboard Alerts Only

C. Additional Catalogs

D. More Workers

Answer: ✅ A. Feature Validation ---

Question 211

## Question 211 A logistics company wants to compare Model V1 and Model V2 before deployment. What is the BEST method?

A. A/B Testing

B. Create More Tables

C. Rename Models

D. Delete Old Version

Answer: ✅ A. A/B Testing ---

Question 212

## Question 212 A fraud model is retrained every night. Management wants automatic approval only if accuracy improves. What is the BEST design?

A. Model Validation Step

B. Manual SQL Query

C. Repartition Data

D. Change Clusters

Answer: ✅ A. Model Validation Step ---

Question 213

## Question 213 A company wants a complete history of model deployments. What should be maintained?

A. Deployment Registry

B. Temporary Views

C. Query Cache

D. Dashboard History

Answer: ✅ A. Deployment Registry ---

Question 214

## Question 214 A marketing model is deployed to production. Engineers need a way to quickly rollback after deployment issues. What is the BEST strategy?

A. Versioned Model Registry

B. Delete Previous Models

C. Disable Tracking

D. Remove Metadata

Answer: ✅ A. Versioned Model Registry ---

Question 215

## Question 215 A recommendation engine uses stale customer activity data. What business risk exists?

A. Reduced Prediction Quality

B. Better Governance

C. Faster Clusters

D. Improved Recall

Answer: ✅ A. Reduced Prediction Quality ---

Question 216

## Question 216 A company wants training and inference features generated using identical business logic. What should be prioritized?

A. Feature Consistency

B. More Partitions

C. Manual Processing

D. Larger Notebooks

Answer: ✅ A. Feature Consistency ---

Question 217

## Question 217 A support team wants alerts whenever model accuracy drops below 90%. What capability should be implemented?

A. Model Monitoring

B. Dashboard Themes

C. Auto Loader

D. Repartitioning

Answer: ✅ A. Model Monitoring ---

Question 218

## Question 218 An ML Engineer wants to understand which dataset version was used during training. What should be tracked?

A. Training Data Lineage

B. Worker Count

C. Notebook Colors

D. Browser Extensions

Answer: ✅ A. Training Data Lineage ---

Question 219

## Question 219 A company trains hundreds of models every month. Management wants standardized deployment procedures. What should be created?

A. MLOps Framework

B. Manual Checklist

C. Shared Spreadsheet

D. Email Process

Answer: ✅ A. MLOps Framework ---

Question 220

## Question 220 A model succeeds technically but produces decisions that violate business policies. What should be added?

A. Business Rule Validation

B. More Training Data

C. Smaller Clusters

D. More Partitions

Answer: ✅ A. Business Rule Validation ---

Question 221

## Question 221 A customer churn model heavily favors one customer group and disadvantages another. What should be investigated?

A. Bias Detection

B. Auto Scaling

C. Data Compression

D. Driver Memory

Answer: ✅ A. Bias Detection ---

Question 222

## Question 222 A retailer wants near real-time product recommendations. What architecture is most appropriate?

A. Batch Recommendation Weekly

B. Real-Time Inference Pipeline

C. Monthly Export

D. CSV Prediction Files

Answer: ✅ B. Real-Time Inference Pipeline ---

Question 223

## Question 223 A company stores prediction results but not input features. Months later they cannot explain model decisions. What key practice was missing?

A. Prediction Traceability

B. Partition Pruning

C. AQE

D. Checkpointing

Answer: ✅ A. Prediction Traceability ---

Question 224

## Question 224 An ML team wants reproducible model training. What should be versioned?

A. Code

B. Data

C. Parameters

D. All of the Above

Answer: ✅ D. All of the Above ---

Question 225

## Question 225 A fraud model generates thousands of predictions daily. Engineers need confidence that production behavior matches training expectations. What should be implemented?

A. Continuous Model Monitoring

B. Less Logging

C. Manual Reviews Only

D. More Catalogs

Answer: ✅ A. Continuous Model Monitoring ---

Question 226

## Question 226 A company wants to prevent deployment of unapproved models. What is the BEST control?

A. Deployment Approval Workflow

B. Manual Naming Standards

C. More Clusters

D. Additional Views

Answer: ✅ A. Deployment Approval Workflow ---

Question 227

## Question 227 A machine learning team discovers duplicate features across multiple projects. What should be created?

A. Central Feature Repository

B. Additional Workspaces

C. More Notebooks

D. More Schemas

Answer: ✅ A. Central Feature Repository ---

Question 228

## Question 228 A model requires customer age, income and transaction history. These are examples of:

A. Features

B. Experiments

C. Workflows

D. Dashboards

Answer: ✅ A. Features ---

Question 229

## Question 229 A prediction service experiences latency spikes. What should be reviewed first?

A. Endpoint Performance Metrics

B. Notebook Theme

C. Workspace Name

D. Catalog Naming

Answer: ✅ A. Endpoint Performance Metrics ---

Question 230

## Question 230 A company wants to compare thousands of experiment runs quickly. What capability is most valuable?

A. Experiment Tracking Dashboard

B. SQL Formatting

C. Table Comments

D. Repartitioning

Answer: ✅ A. Experiment Tracking Dashboard --- --- title: Databricks Certification Practice Test 11 description: AI, GenAI and Advanced ML Scenario Questions ---