Framework + Systems (Guide)

Reference extracted from the detailed guide.

Framework (Memorize This)

1. CLARIFY (5 min)
   - What's the business goal?
   - What are the constraints? (latency, scale, cost)
   - What data is available?

2. METRICS (5 min)
   - Offline: Precision, Recall, AUC, NDCG
   - Online: CTR, conversion, engagement, revenue

3. DATA & FEATURES (10 min)
   - Data sources
   - Feature engineering
   - Feature stores

4. MODEL (10 min)
   - Model selection
   - Architecture
   - Training strategy

5. SERVING (10 min)
   - Inference pipeline
   - Latency optimization
   - Scaling

6. MONITORING (5 min)
   - Data drift
   - Model degradation
   - A/B testing

Must-Study Systems

1. Recommendation System (Netflix/YouTube)

Video: https://www.youtube.com/watch?v=n3RKsY2H-NE

Key Components:

• Candidate generation (recall)
• Ranking (precision)
• Two-tower model
• User/item embeddings

2. Feed Ranking (Facebook/Twitter)

Video: https://www.youtube.com/watch?v=hKoJgLf5sj0

Key Components:

• Multi-stage ranking
• Real-time features
• Personalization
• Diversity/freshness

3. Ads Click Prediction

Video: https://www.youtube.com/watch?v=RZJBOo9HW-M

Key Components:

• CTR prediction
• Wide & Deep model
• Feature crosses
• Calibration

4. Search Ranking

Key Components:

• Query understanding
• Document retrieval
• Learning to rank
• BERT for search

Book

"Designing Machine Learning Systems" by Chip Huyen

• Buy: https://www.amazon.com/Designing-Machine-Learning-Systems-Production-Ready/dp/1098107969
• This is the gold standard for ML system design

---