This document discusses building a generic text classifier to identify spam accounts on social media platforms like Twitter. It provides statistics on spam accounts and mobile users, describes the data pipeline used for data acquisition, cleaning, feature engineering and text classification. The classifier achieved an AUC of 0.91 when combining Twitter data and SMS data and a mean accuracy of 0.70 using just meta features with a random forest classifier. Next steps include incorporating more data sources and meta features and providing the classifier as an API.