Using MongoDB for the Art Genome Project

Daniel Doubrovkine

Head of Engineering at

October 3, 2011 is a fine art search engine that uses the Art Genome Project to infer connections between tens of thousands of works by classic and contemporary... artists. After building a prototype with MySQL it became clear that a relational database was poorly suited for storing data in a high dimensional space. We have then adopted MongoDB and are now running it as our only database in production, searching a 560-dimensional genome space in under 1/4s per query using RoR/Mongoid.