ML Interview Checklist: What to Review Before Your Next Round

ML interviews usually mix theory, practical ML, and systems thinking. Here’s a compact checklist that covers what you’re most likely to be asked.

1) Metrics and evaluation

• classification: precision/recall, ROC-AUC, PR-AUC
• regression: MAE/RMSE, outliers
• ranking: NDCG, MAP
• calibration: why it matters

2) Bias/variance and generalization

• underfitting vs overfitting
• regularization (L1/L2), early stopping
• cross-validation strategy

3) Feature engineering and leakage

• leakage examples (time-based splits!)
• categorical encoding tradeoffs
• missing values strategy

4) Experimentation

• offline vs online metrics
• A/B test basics (power, duration, guardrails)

5) ML System Design (what interviewers want)

• data collection and labeling pipeline
• training and serving architecture
• monitoring: drift, quality, latency, costs

Quick answer pattern

When asked “How would you build X?” answer:

1. goal + metric
2. data sources
3. baseline model
4. iteration plan
5. monitoring + rollback