A Large Scale Content Ranking Platform as Applied to E-commerce Store Fronts

An e-commerce marketplace like Flipkart displays hundreds of millions of promotional content in a day on its home page. To display content, we predict the probabilities of various responses by a user for any given content. In this paper, we describe the various considerations and challenges in building a ranking platform for these content at Flipkart. We describe the architecture and delve into the solutions used in practice, in modelling, training, scaling, data correctness, freshness and metrics to provide a comprehensive overview of such a system.