How do you design and implement a recommendation system?

Recommendation systems predict user preferences using collaborative filtering, content-based, or hybrid approaches.\n\n• Collaborative filtering: User-item matrix factorization, neighborhood methods\n• Content-based: Feature similarity, TF-IDF, embeddings\n• Hybrid: Combine multiple approaches, ensemble methods\n• Evaluation: Precision@K, NDCG, diversity metrics, A/B testing\n\nExample: Netflix uses matrix factorization for collaborative filtering, content features for cold start, and deep learning for sequential recommendations. Handle sparsity with regularization, use implicit feedback, and optimize for business metrics.