Amazon Personalization
May 2025 – August 2025- Designed a configurable, bulk data ingestion pipeline that reduced the time for developers to onboard new ML model features from 6 weeks to one day.
- Developed a low-latency online featurization component that allows developers to configure which features to use during inference.
- Utilized AWS Glue to perform offline ETL.
- Conducted feature engineering experiments by introducing a 158% increase in features, performed feature importance analysis and distribution studies, and evaluated model performance through key metrics.

