1. Past research experirence (reinforcenment learning, model explainability). A lot of followups. 2. Bias and variance trade-off in random forest. 3. Linear regression closed-form solution (w/ and w/o L2 regularization).