Search Ranking System

Design a machine learning-based search ranking system for an e-commerce platform.

Scenario: An e-commerce site with 10M products needs to rank search results for 100K queries per second. Results must be relevant, personalized, and consider business objectives (margin, inventory).

Your Task: Design the complete search ranking pipeline.

Key Challenges:

1. Two-stage architecture (retrieval + ranking)
2. Learning to rank from click-through data
3. Position bias in training data
4. Balancing relevance with business objectives
5. Real-time personalization