Product Feed and Shopping Recommendation System

Design a recommendation system for a product feed, Jira-style recommendations, or an Amazon-style shopping surface. Variants distinguish landing-page recommendations from search-engine recommendations and expect discussion of ranking, retrieval, feedback, and latency. Requirements Generate recommendations for a product feed, cross-product Jira/Confluence surface, or shopping experience. Separate candidate generation from ranking. Consider landing-page recommendations and search-result recommendations separately when the prompt asks for both. Incorporate user signals, item/content metadata, and recent interactions. Discuss latency constraints for online ranking and batch/offline pipelines for model training or feature generation. Define metrics for relevance, engagement, freshness, and business quality. Notes MLE candidates may receive this as a relatively standard ML system-design prompt, but interviewers can still go deep on product-specific retrieval and ranking choices. Clarify whether recommendations must be explainable, permission-aware, or scoped to a specific Atlassian product. Preparation Sketch retrieval, ranking, feature store, experimentation, and monitoring components. Prepare trade-offs for collaborative filtering, content-based retrieval, two-tower retrieval, and business-rule reranking.