System design for ML engineers: how different is the interview?

i'm transitioning more into ML infra and recently had a system design interview that felt significantly different from traditional backend ones. for ML roles, how much traditional system design depth is actually expected versus the ML-specific parts? i was asked to design a personalized recommendation system. i spent a lot of time on model serving infrastructure, feature stores, online/offline pipelines, and data versioning. but then the interviewer also dug into general distributed system concerns like fault tolerance for the serving layer, API design, and database choices for the metadata store. it felt like two interviews in one. i'm wondering if there's a common split or if it varies a lot by company what they emphasize for ML system design.