Currently, most recommendation systems make predictions of a user’s “preferences” based on (1) similarity across users and (2) their preferences. While these recommendation engines prove effective for items such as movies and books, they work poorly in regard to items that are used in combination with other items. Examples include: clothing items, furniture, grocery items, medicines, etc. Recommending these types of items requires a combination-based recommendation system that, given an item, can suggest valid combinations containing that item. This problem is challenging because (1) features are difficult to identify, (2) quality as a function of all items in the combination is difficult to compute, and (3) the number of possible combinations can be exponential.

This SuperUROP project will serve as a continuation of a startup, Asorti, I have been working on since February, 2013 with my two co-founders, Manasi Vartak (Ph.D. student at MIT), and Lauren Clark (management student at MIT). This past April, Asorti was named finalist for DreamIt (a NYC-based startup accelerator) and Y Combinator (a CAbased accelerator).