App ads serving

Worked on the Platform team, whose primary role to delivery high quality web application product for our tenants and customers.

Worked on the Reliability team, whose primary role is to create internal tooling to monitor and audit the platform.

Working on the Account Manager Digital Assistant System based on Machine Learning, Natural Language Processing, NiFi, and Hive.

Tracking number email request classification project:

Added a whole new functionality to Amanda System to reduce up to 15% daily workload for Account Managers.

Designed a process algorithm to automatically distinguish and reply tracking number request emails.

Trained a Random Forest model and polish the performance based on feature extraction strategies, variables adjustment, and results visualization.

Designed a multi-cases Regex procedure based on email analysis, pre-processing and text mining.

Built a pre-filter system which significantly saves up to 30% run-time.

Tested and improved performance for Regex model and Random Forest model based on cross validation, random shuffle split, and confusion matrix.

Built a tracking number verify&correct system based on Regex, HTTP query, and database match.

Classification performance reaches over 97% accuracy and has low latency in real time environment.