People You May Know System

Design a People You May Know feature that proposes relevant new connections from a social graph. This written exercise builds on the feed and movie-recommendation chapters. Use a hypothetical platform and state your assumptions; the example does not describe a particular company’s implementation.

Frame the recommendation task

Decide what a successful suggestion means before choosing a model. Is the objective a connection request, an accepted request, or a connection that leads to later interaction? Those targets require different labels. Also define which accounts are eligible to appear and how existing connections are excluded.

For a 60-minute practice session, reserve time for scope, candidate generation, ranking, evaluation, and a final walkthrough. Adjust the allocation as you work so you can explain a complete design.

Build candidates from the graph

Start with a small example. Suppose A is connected to B and C, and both B and C are connected to D. D is a candidate connection for A because the graph supplies two shared connections. That is evidence for retrieval, not proof that A knows D or wants the suggestion.