Movie Recommendation Service System

Instructors: Christian Kaestner and Eunsuk Kang

Course Website

Project summary

• Functioned as a machine learning engineer in a group to create and maintain a movie recommendation service for ∼1 million users and 27,000 movies, ranked the first in class for service response time
• Implemented a hybrid recommendation system that combined collaborative filtering, content-based, and rule-based approaches, resulting in high-quality recommendations (top 15% in class) while maintaining low latency
• Devised evaluation strategies and metrics for service operations, including online/offline model assessment, A/B testing, provenance tracking, drift detection, and fairness measurement