Amazon - ML

Designed machine learning framework for continuous training and improvement of real-time inference models using ground-truth labels and freshly generated feature sets

Implemented automated model rollbacks and data quality checks to preemptively revert and alarm on model performance deviation

Architected generic data lake for running ETL and analysis jobs on mixed data sets from machine learning models along with A/B testing capabilities

Productionized NLP/CV machine learning models for usage in high-scale recommendation systems and added distributed batch computation capability for prediction generation workflows