In this tutorial, we will present Nubank's newest open sourced library, FKLearn. You will learn how we combined the best practices from functional programming with the most powerful Machine Learning libraries to empower data-driven solutions at scale. We will build a full machine learning model to predict spending behavior on credit card using FKLearn main tools.
Nubank is a Brazilian technology company that provides financial services. Nubank strongly relies on data-driven decision making and Machine Learning to provide the best services for our customers and FKLearn is our main library that allows us to quickly create and deploy machine learning that support key company areas such as credit risk, and customer service. Nubank was elected the most innovative company of 2018 in Latim America by Fast Company Magazine and we strongly believe in sharing our technological knowledge with the community. We currently have the biggest data science Meet-up in Latin America and we want to share some of our developments in machine learning with the Latin America Python community.
FKLearn is a fully functional python library for Machine Learning.
The three pillars for FKLearn are:
Reproducibility: We want to guarantee that every previous trained model can be reproduced and reevaluated.
Deployability: We want to be fast at deploying new models and improvements on existing models.
Real-life validation metrics: We want to guarantee that our model are fully validated and we are safe that the models don't have blindspots.
On this tutorial we will show how functional programming guarantees that our machine learning models can be easily reproduced and deployed in production and also go into details of how real-life validation of models works and how FKLearn provides an easy to use interface for validation methods.