Online feed ranking updates
Design an online (continual) learning system for a feed-ranking model that must incorporate fresh engagement signal within minutes — a video going viral, a breaking news event — rather than waiting for a daily batch retrain. The risk: continually updating a live model on streaming feedback can let a feedback loop, a data spike, or a bad batch corrupt the model in production. Walk through the streaming-update architecture, how you push updated weights to the serving fleet safely, and how you contain or roll back a bad update.
What a strong answer looks like
Clarify scale and constraints first. Propose a clean component breakdown, then go deep on the hard parts (data model, bottlenecks, consistency, failure modes) and name the trade-offs you are making.
Clarify5:30 left
Estimate5:30 planned
Design16:30 planned
Deep dive13:30 planned
Failure9:00 planned
Which questions mattered is sealed until you submit. Telling you now would just be handing over the edge cases.
Run or narrate your approach, then ask the coach.