This document proposes a personalized review engine called Yelper Helper that consists of two parts: 1) predicting the usefulness of new reviews using a zero-inflated Poisson regression model on user, business, and review attributes from the Yelp dataset; and 2) computing user similarity through collaborative filtering to uncover taste preferences using a use-taste matrix and restaurant-category matrix. The goal is to address the problem that around 75% of Yelp reviews have zero useful tags by taking different modeling approaches to address issues in the data for similarity and prediction.