Engineered and integrated signals from manual URL reviews & flagging rules into a deep neural network

The deep neural network was one that attempted to predict if a certain site was safe to host ads on based on Google AdSense policies

Through experimentation and data analysis, I was able to increase the area under the Precision-Recall curve (used as a metric for performance) from 0.285 - 0.354

Constructed, unit-tested, and integration-tested a pipeline to reformat manual URL reviews such that they could be easily joined to one of Google Ads' central data structures; this data structure in particular is used for monitoring publisher behavior and content