Design a Feed Generator

Design a high-scale content discovery engine with ranking and personalization.

What is the Feed Generator Problem?

Design a feed generator system that aggregates content from multiple sources, ranks posts based on various algorithms, and delivers personalized feeds to users. The system should support different ranking strategies, caching mechanisms, real-time feed updates, and content filtering. The design should be extensible, performant, and handle high volumes of concurrent requests.

In this problem, you’ll design a service that builds a custom view for every user, balancing freshness with relevance while handling millions of concurrent content updates.

Problem Overview

Design a discovery engine that collects posts from a user’s network, ranks them based on interest and quality, and serves a smooth, personalized experience.

Core Requirements

Functional Requirements:

Aggregation: Collect content from followed users, pages, and groups.
Personalization: Rank posts using multiple strategies (Latest, Popular, Relevant).
Filtering: Remove duplicate, low-quality, or sensitive content.
Real-time Updates: Notify users of new relevant content as it arrives.
Multi-content Support: Handle text, images, videos, and shared links.
Social Metrics: Integrate engagement (likes, comments) into the ranking logic.

Non-Functional Requirements:

Ultra-Low Latency: Feeds must load in milliseconds even with millions of posts.
Consistency: Ensure users don’t see the same post twice in a single session.
Scalability: Handle massive fan-out (one celebrity post going to millions of followers).
Availability: The system must remain fast even during traffic spikes.

What’s Expected?

1. System Architecture

The generator sits between the ContentStore and the UserFeed, coordinating between the Ranker and Cache.

2. Key Classes to Design

classDiagram
    class FeedGenerator {
        -RankingStrategy ranking
        -List~Filter~ filters
        -FeedCache cache
        +generateFeed(user)
        +onNewPost(post)
    }
    
    class RankingStrategy {
        <<interface>>
        +score(post, user) double
    }
    
    class Post {
        -String id
        -User author
        -ContentType type
        -EngagementStats stats
    }

    class FeedCache {
        -Map~UserId, List~Post~~ storage
        +invalidate(user)
        +update(user, post)
    }

    FeedGenerator --> RankingStrategy
    FeedGenerator --> FeedCache
    Post --> ContentType

System Flow

Personalized Ranking Flow

Key Design Challenges

1. Ranking Latency

Sorting 10,000 posts for every “refresh” click is too slow.

Solution: Use Asynchronous Pre-computation. When a post is created, the system “pushes” it into the pre-computed feeds of active followers. For inactive users, the system only computes the feed “lazily” when they log in.

2. The “Celebrity” Fan-out Problem

When a user with 50 million followers posts, updating 50 million caches simultaneously will crash the system.

Solution: Hybrid Model (Push + Pull). Push posts from “Normal” users to their few followers. For “Celebrities,” don’t push. Instead, when a follower of a celebrity loads their feed, the system “pulls” the celebrity’s recent posts and merges them into the ranking on-the-fly.

3. Dynamic Ranking Rules

Business wants to promote “Video” content this week and “Breaking News” next week.

Solution: Use the Strategy Pattern combined with a Weighted Scorer. The RankingStrategy iterates through a list of Scorer components (RecencyScorer, EngagementScorer, TypeScorer). You can change the “weight” of each scorer via a configuration file without changing the code.

What You’ll Learn

By solving this problem, you’ll master:

✅ Feed Optimization - Balancing Push vs. Pull models.
✅ Personalization Logic - Building flexible weighted ranking systems.
✅ Cache Invalidation - Managing user-specific snapshots at scale.
✅ Event-Driven Architecture - Handling massive data fan-outs gracefully.

View Complete Solution & Practice

Ready to see the full implementation? Open the interactive playground to access:

🎯 Step-by-step guidance through the 8-step LLD approach
📊 Interactive UML builder to visualize your design
💻 Complete Code Solutions in Python, Java, C++, TypeScript, JavaScript, C#
🤖 AI-powered review of your design and code

🚀 View Solution in Playground

After mastering the Feed Generator, try these similar problems:

Search Index - Building high-speed retrieval systems.
Notification Service - Handling massive broadcast events.
Chat Room Manager - Managing real-time message streams.