Building A Large Scale Recommendation System People You May Know

LinkedIn’s connection recommender makes a billion-person network manageable by narrowing the candidate pool before applying its most expensive predictions. Initial searches find a few thousand potential connections through network relationships, similarity and simple rules. A lightweight ranking stage reduces those candidates to hundreds, then deeper models estimate outcomes such as whether an invitation will be sent and accepted. A final stage adjusts the recommendations for diversity and fairness. This funnel concentrates computation on plausible connections while allowing the final list to balance goals beyond predicted engagement.