Q1
Walk me through how you would approach feature engineering for a high-dimensional dataset with 500+ variables. What techniques would you use to reduce dimensionality and why?
Why they ask this:* They want to assess your understanding of practical ML workflows, feature selection methods (PCA, correlation analysis, domain expertise), and ability to balance model complexity with performance.
Q2
Explain the difference between regularization techniques (L1 vs L2) and describe a scenario where you'd choose one over the other in a production model.
Why they ask this:* This tests your grasp of overfitting prevention, model interpretability, and ability to make trade-offs between bias and variance in real-world applications.
Q3
You're working with imbalanced data where one class represents 95% of your dataset. How would you handle this problem, and what metrics would you use to evaluate model performance?
Why they ask this:* They're evaluating whether you understand class imbalance pitfalls, techniques like SMOTE or stratified sampling, and why accuracy alone is misleading—critical for practical data science.
Q4
Describe your experience with SQL and data pipelines. How would you optimize a slow query that joins three large tables with millions of rows?