36,024 views
Learn ML with us: ML Simulator: http://bit.ly/3L8JT1c Start ML Course: http://bit.ly/3T9vrs0 Hard ML Course: http://bit.ly/41ZdhwP The hero of the new interview for the position of Junior ML engineer was Vadim, a graduate of Start ML and participant of the ML Simulator. Vadim had no experience in IT before the karpov.courses programs. We invited him to take part in a mock interview and find out how ready he was to enter a new profession. And the interview was conducted by Bogdan Pechenkin, the author of the Simulator. Vadim's LinkedIn: / vadimbaev 0:00 Introduction and acquaintance 2:55 Uplift-modeling 5:50 Churn model 08:06 Metrics for class imbalance 11:32 Calculating the confidence interval of the area under the PR curve 12:35 How the bootstrap works 15:26 Bootstrap feedback 16:52 Estimating the metric if there is no possibility to take samples, but there are trained models and their estimates 19:00 Cross-validation schemes 20:55 The influence of bias and dispersion on the quality of the model 22:43 Main components of the A/B test 25:42 P-value and hypotheses 27:27 The required amount of data 29:31 Running simulations 32:40 Gradient boosting 35:00 What happens if you remove the first tree in gradient boosting 36:45 Differences in tree construction 40:10 Evaluation of regression model predictions 43:48 Deploy ML services 47:20 Web frameworks 48:13 Fixing the version of libraries 50:00 What metrics can be monitored after deploy 51:33 What to do if the model did not calculate the prediction, and the answer is needed quickly 53:12 Modeling demand forecast 55:36 Practice 1:06:15 Feedback