Chelsea Ju

Experience

1. Applied Scientist Intern

   Amazon

   Sep 2019 — Present

   Led and implemented the machine learning infrastructure design to support ad-hoc experiments on customer data for the trained ML models, and to improve the efficiency of model training/evaluation cycle.* Constructed and deployed a quantized speaker verification (SV) model to edge devices.* Trained and evaluated deep learning based (RNN) text-independent SV models across different emotion spectrum.* Implemented a confidence model to reduce accuracy variability among sub-populations for SV system.* Mentored two summer intern projects; one leverages CycleGAN to synthesize emotional speech, and the other applies adversarial training to improve speaker identification model.* Performed large-scale data analysis in a high-throughput parallel computing system (AWS) using PySpark.