DescriptionOnline advertising is able to target users at a fine level of granularity. To do this effectively, models are required to represent users and their behavior. In this thesis, we studied several problems related to models of online users. In online advertising, advertisers and ad platforms use user profiles as the language to target users, composed of user information from demographics, location, and interests. We implemented a user-profile-driven ad crawling framework and empirically investigated the relationship between user profiles and the ads to which they are exposed. We observed user profiles to play a greater role in display ads than in video ads. Furthermore, the main mode of accessing online content has been shifting from website browsing to mobile application usage. Mobile apps have become the building blocks to model user behavior on mobile devices. We designed a neural network model (app2vec) to vectorize mobile apps by studying how users employ these apps. We analyzed the learned app vectors qualitatively and quantitatively and used them to extract user app usage profiles for app-install advertising. Finally, advertisers are faced with the challenge of finding the optimal user profile properties to target. We designed a look-alike audience extension system, where advertisers provide a list of past converters as "seed users" and our system determines users similar to the seed. Rather than assuming linear separability of lookalike and non-lookalike users, as in prior work, we propose a new approach with nearest-neighbor filtering. Our system works efficiently for billions of users and improves the ad campaign conversion rate in practice at Yahoo!.