Whenever you're ready, walk me through how you'd approach this.
How would you design a chatbox system similar to Facebook?
Clarify whether the scope includes real-time message delivery, conversation-list ranking, suggested replies, spam detection, or all of these. Explain the end-to-end ML architecture, including data collection, candidate generation, ranking, online versus batch inference, latency and availability targets, evaluation, monitoring, and failure handling. State assumptions explicitly where the question is ambiguous.