Deep-Dive Expect to justify every decision you made. Focus on the "Why" rather than just the "What." • Problem Statement: What specific business or research gap were you filling? • Data Pipeline: How did you handle missing values, outliers, or imbalanced classes? • Model Selection: Why choose XGBoost over a Random Forest or a Neural Network for this specific data? • Evaluation: Which metrics mattered most (e.g., Precision vs. Recall) and why? 2. Basic ML Algorithms You should be able to explain these as if speaking to a non-technical stakeholder, then dive into the math. • Linear/Logistic Regression: Understanding assumptions (linearity, homoscedasticity) and regularization (L_1 and L_2). • Decision Trees & Ensembles: How entropy/Gini impurity works; the difference between Bagging and Boosting. • K-Nearest Neighbors (KNN): The impact of distance metrics and the "curse of dimensionality." 3. Probability & Statistics This is the "engine" under the hood of ML. Common topics include: • Distributions: Knowing when to use Normal, Bernoulli, or Poisson distributions. • Bayes' Theorem: Calculating conditional probability (often framed as a disease testing word problem). • Hypothesis Testing: Defining p-values, Type I/II errors, and A/B testing logic. • Central Limit Theorem: Why it allows us to assume normality in large samples.