Meta (Facebook) Interview Guide: Process & Questions

Approach: BFS with column tracking. Algorithm Steps: 1. Assign column 0 to the root 2. Left children get column - 1, right children get column + 1 3. Use a hash map (column -> list of values) and a queue of (node, column) pairs 4. Process level by level (BFS ensures top-to-bottom order within each column) 5. Sort the keys of the hash map for left-to-right order Complexity: • Time: O(n log n) due to sorting columns • Space: O(n) for the hash map and queue Edge Cases to Handle: • Empty tree (return empty list) • Skewed tree (all nodes in same column) • Nodes at same row and column (maintain insertion order) Meta-Specific Tip: Write the solution first, then walk through a small example to verify. Code clarity matters — use descriptive variable names and handle the null-root case upfront. Meta interviewers want clean, runnable code, not pseudocode.

Two Fundamental Approaches: 1. Push (fan-out on write): When a user posts, immediately write to all followers' feed caches • Works for users with few followers • Breaks at celebrity scale (100M followers = 100M writes per post) 2. Pull (fan-out on read): When a user opens their feed, fetch posts from all people they follow and rank in real-time • Expensive for users following many people Hybrid Approach (What Meta Actually Uses): • Push for normal users (most have <1000 followers) • Pull for celebrities/high-follower accounts Ranking Pipeline: • ML model scores each candidate post on relevance • Factors: engagement history, recency, content type, social closeness • The ranking service takes candidate posts and returns a scored, ordered list Storage Architecture: • Posts: distributed database (like TAO) • Feed cache: distributed key-value store (like Memcached) • Media: CDN with edge caching • Feed rendering service stitches together post data, user profiles, and media URLs