At OLX, many anti-fraud Machine Learning models need online features, transformed and aggregated in real time and served to the model as fast as possible. Data Scientists create offline features with data from our Datalake and they do it with queries, but they depend on Data Engineers when creating online features. Creating online features should be an easy task to a Data Scientist. In this session, we’ll present a feature store where Data Scientists can create their online feature using queries, using KSQLdb as engine transformation.
