NewsBreak Interview Questions

Four of the five NewsBreak questions recorded here come from Machine Learning Engineer interviews, with a single Software Engineer report alongside them. The notable pattern is that all five were asked in a technical screen; no onsite round has been reported for this company yet, so the screen is carrying the whole filter. Three are modelling questions and they go past definitions. One asks how you would approach class imbalance, which means arguing resampling against class weights against threshold moving, and saying what each does to precision and recall on the minority class. Another asks you to derive temperature scaling and then implement it, so you need to know that one scalar on the logits repairs calibration without moving the argmax or the ranking. A third wants the parameters of an optimal piecewise constant regression, which is a dynamic programming problem wearing statistics clothes. The remaining two are conversational: walking an interviewer through a project you actually owned, and explaining why this company and why a startup rather than somewhere larger. Four are rated medium and one hard. Every question carries a written solution, none opens a console, and 80% of the set is readable without premium.