IDEAS SoCal Data Science Conference 2017

The mission of Tala is to provide financial access, choice, and control to the unbanked or under-banked population in the developing world. Our primary product is a short-term loan applied for through our mobile app. In the developing world, we face the challenge that almost none of our applicants have any traditional credit bureau data. Instead, we use a wide variety of behavioral data from their mobile phones with machine learning to underwrite our customers.

In this talk, we'll discuss some of the interesting feature engineerings we use to predict credit performance, as well as a few challenges our data science team has faced. One challenge we've faced is the "cold start" problem when we enter a new market. In this talk, I'll discuss a transfer learning approach to addressing this challenge. Second, we'll discuss some of the microeconomic and macroeconomic insights we can find from this dataset with the recent Kenyan election as an example. Finally, we'll discuss some of the ethics involved in using personal mobile data to make underwriting decisions.