Research work on graph-based learning where the structure between examples matters as much as the labels themselves.

This sits on the site as proof of fundamentals: modeling, experiments, data representation, and the patience to work through less obvious ML problems.