Nowadays data science has been playing a vital role in the IT industry, as it has been used to inform the actual business decisions at many IT companies. At Snap, we use data to understand and gain insights on how our users are using our products and make day-to-day business decisions accordingly. Because of the dynamics, heterogeneity and tremendous volume of the real user data, there are many new challenges for data science research at the industrial-scale. In this talk, we share a few examples of the data science research effort at Snap. In particular, we showcase a piece of recent work that proposes a framework on clustering millions of Snapchat user into a few interpretable types based on their intention of using the app, and at the same time, maximally preserves user privacy. From these examples, we show how data science applications in practices differ from theories, and how we tackle real-world data science challenges at scale.