Machine Learning Engineer Interview Questions

Machine learning engineer loops split almost evenly between two stages: technical screens supply 423 of the 907 reported questions and onsites 420, with online assessments at 40 and HR screens at 15. Coding and algorithms is still the largest category at 321 questions, and the samples show those problems in technical screens, so the first stage often looks like a software engineering screen. ML-specific weight sits just behind: 218 machine learning questions and 196 ML system design questions, which the samples show clustering in onsites. Meta, OpenAI and Amazon report the most, followed by Google, Pinterest, Microsoft and Apple. Most reports carry no level tag; 88 are Senior+, 19 intern and 18 new grad. Concrete themes recur. ML system design questions ask for end-to-end products: a sequential personalized playlist recommender, a model predicting next-week streaming engagement, an embedding and classification API. Modeling questions push on why rather than what: vectorizing one-nearest-neighbor and expressing it as a neural forward pass, or diagnosing a production classifier and defending its systems design. Coding rounds stay standard, asking for the median of an integer stream or a linked list reversed in groups of k. Statistics appears too, such as explaining why a geometric mean can exceed its median. 101 first-hand interview experiences accompany the questions.