Design Apple News without ML
Quick Overview
This question evaluates a candidate's ability to design search and content-discovery architectures for a news application without a trained ranking model, assessing competencies in information retrieval, metadata extraction and indexing, rule-based ranking and heuristics, freshness and editorial signals, personalization heuristics, scalability, latency, and logging/evaluation. It is commonly asked to probe systems-design thinking and trade-offs between relevance, recency, safety, and operational constraints within the domains of system design and information retrieval, and it focuses on both conceptual architectural reasoning and practical implementation considerations.
Design an initial search and content-discovery system for a news application similar to Apple News, assuming you do not have a trained ranking model yet. Users should be able to search for articles, topics, and publishers, and the system should surface results that are relevant, fresh, safe, and reasonably personalized.
Discuss:
- article ingestion and metadata extraction,
- indexing and storage,
- rule-based or heuristic ranking,
- freshness handling,
- topic and publisher filtering,
- editorial signals,
- scalability and latency,
- logging and evaluation,
- and how you would later evolve the system toward ML-based ranking.