Search Engine / FB Post Search
Difficulty: Hard
What It Tests
Search index building, relevance ranking, freshness, query parsing at social-network scale.
Topics Covered
- Search & Elasticsearch
- Stream Processing
- Big Data Architectures
- Database Design: SQL vs NoSQL
- Database Indexing
Hello Interview Breakdown
Read the full Hello Interview breakdown →
Approach Hints
- CDC (Debezium) captures new/updated posts from primary DB and publishes to Kafka; indexing workers consume and upsert Elasticsearch documents
- BM25 relevance scoring augmented with personalization signals (friend graph proximity, engagement history)
- Near-real-time indexing via Elasticsearch's 1-second refresh interval; use a hot index for recent posts and a cold index for historical
- Query parsing handles hashtags, mentions, phrases, and boolean operators; a ranking model blends text relevance + social signals