AI and Machine Learning Fundamentals

⏷ 6. Why do we split data into training, validation, and test sets?

⏷ 7. What is data leakage, and how do you prevent it?

⏷ 8. What makes a dataset representative?

⏷ 9. How do you handle an imbalanced classification dataset?

⏷ 10. How should you handle missing, outlier, and categorical feature values?

⏷ 11. What are underfitting, overfitting, and generalization?

⏷ 12. How do a loss function and an evaluation metric differ?

⏷ 13. How do gradient descent and the learning rate work?

⏷ 14. How do regularization and early stopping reduce overfitting?

⏷ 15. How do model parameters and hyperparameters differ?

⏷ 16. How do you read a confusion matrix and choose between precision, recall, and F1?

⏷ 17. How do you choose a classification threshold?

⏷ 18. When should you use ROC-AUC versus PR-AUC?

⏷ 19. How do MAE, MSE, RMSE, and R-squared differ?

⏷ 20. Why do you need a baseline, and when is cross-validation useful?

⏷ 21. How do batch and online inference differ?

⏷ 22. How do data drift, concept drift, and training-serving skew differ?

⏷ 23. What should you monitor for a production ML model?

⏷ 24. How do feedback loops and delayed labels affect an ML system?

⏷ 25. How do you deploy a new model version safely?

⏷ 26. When should you prefer a simple model over a more complex one?

⏷ 27. How do you decide between building a model and using a pretrained or managed model?

⏷ 28. How do explainability, calibration, and fairness affect model design?

⏷ 29. How would you design an end-to-end machine-learning feature pipeline?

⏷ 30. How would you debug a model that performs well offline but poorly in production?

⏷ 31. How does probability calibration differ from discrimination, and when does it matter?

⏷ 32. How do feature stores reduce training-serving skew, and what problems do they not solve?