Knots | INFORMS

A lot of social recommendation systems reinforce existing clusters. We wanted to see whether software could deliberately create more weak ties between students.

Students could coordinate meals in real time, while the recommendation system used interests, compatibility, and graph structure to suggest people outside their usual cluster.

I led the team and worked full stack across the Flutter client, TypeScript/Firebase backend, location features, notifications, and the matching logic.

The project turned a graph-theory idea into a product question: can an algorithm move beyond recommending a profile and actually get two people who would not normally meet into the same room?