← Back to categories

Debugging & Failure Stories

Model Performance Debugging

8 questions

Multiple Choice

Diagnosing Model Performance Drop

Interpreting Confusion Matrix Patterns

Feature Importance Debugging

Debugging Inconsistent Cross-Validation Results

Open-Ended

Systematic Debugging of Poor Model Performance

Diagnosing and Fixing Overfitting

Diagnosing and Fixing Underfitting

Debugging Feature Engineering Issues

Data Quality Issues

8 questions

Multiple Choice

Detecting Feature Encoding Errors

Identifying Label Noise Impact

Data Pipeline Debugging

Missing Value Strategy Selection

Open-Ended

Hunting for Data Bugs

Detecting Feature Corruption

Debugging Label Quality Issues

Debugging Data Pipelines

Training & Convergence Issues

8 questions

Multiple Choice

Diagnosing NaN Loss Values

Loss Plateau Diagnosis

Unstable Training Loss Pattern

Vanishing Gradients Identification

Open-Ended

Model Not Converging

Debugging Vanishing and Exploding Gradients

Interpreting Loss Curves

Debugging Hyperparameter Problems

Leakage & Contamination

8 questions

Multiple Choice

Identifying Target Leakage

Train-Test Contamination Detection

Temporal Leakage in Time Series

Feature Leakage from Aggregations

Open-Ended

Detecting Target Leakage

Train-Test Contamination

Temporal Leakage in Time Series

Subtle Feature Leakage Patterns

Production Failures

8 questions

Multiple Choice

Incident Response Priority

Production Rollback Decision

Diagnosing Production Model Drift

Postmortem Analysis Focus

Open-Ended

ML Incident Response

Model Rollback Decisions

ML Postmortem Analysis

Debugging Model Degradation in Production

Silent Failures

8 questions

Multiple Choice

Metric-Target Misalignment

Detecting Silent Model Degradation

Subtle Data Pipeline Bug

Identifying Feedback Loop Issues

Open-Ended

Finding Subtle ML Bugs

Offline-Online Metric Mismatch

Detecting and Handling Distribution Shift

Feedback Loops in ML Systems