Nice, I actually read that Jina article when it was published, but forgot they use facility location as well! The saturated coverage algorithm looks pretty interesting, I'll have a look at how feasible it would be to add that to Pyversity.