Social network Pinterest has purchased a startup that powers a machine-learning recommendation engine. The seven-person company Kosei, which was founded less than a year ago, has developed a graph that understands the relationships between millions of products. Image-driven Pinterest can use this recommendation engine to help detect spam, recommend more relevant pins, and target its advertising. Pinterest just rolled out advertising on its platform earlier this month.

With the help of the Kosei team, Pinterest can bolster its machine learning initiatives. Founders Lance Riedel and Jure Leskovec are both experts in machine learning. According to Pinterest’s company blog, almost every team from the Monetization team to the Black Ops team are using machine learning. For example, the Visual Discovery team is working on algorithms to improve object recognition on the image-heavy site.

A recommendation engine powered by machine learning is intended to keep users on the social network longer. Michael Lopp, Pinterest’s head of engineering, told The Wall Street Journal that he wants to see new users be able to discover content that interests them from their first moment on Pinterest. The company wants to be able to recommend boards as well as pins.

Some of those recommended pins may even be advertisements. Kosei’s engine understands more than 400 million relationships between 30 million products. They set out to “empower commerce” and they will certainly get that chance with Pinterest’s newly established advertising system.

Pinterest hopes the Kosei acquisition will help with ad targeting as the social network will continue to improve its conversion-tracking and offline sales.