Research & Development for Sberbank

In this commercial project, we work with Sberbank — the largest bank in Russia — to deploy machine learning methods to find and fulfill people needs.

One of the main challenges is to engineer feature-representations for the bank’s customers. We feed these representations into clustering, visualization, and topic analysis tools to get handy visualizations of the people preferences, and how do they evolve over time. The goal is to build a tool that automatically understands which broad and narrow people groups are there and also can target a particular category of people needed by the bank. For example, we may look for individuals who save money on their account but do not receive any interest on that, because they lack the understanding of the deposit options available.