Online Movie Recommender System This is an independent project. Written in Python, Tensorflow. Preprocessed raw data into Spark Dataset by implementing MapReduce operations on MovieLens dataset. Retrieval: Built candidate generators by implementing content-based filtering and collaborative filtering via Matrix Factorization using SparkMLib, visualized item / query embedding by using PCA, experimented with different similarity measures and regularization methods. Ranking: Ranked candidates by implementing Factorization Machines and Gradient Boosting Decision Tree + Logistic Regression using SparkMLib. Re-ranking: Experimented with factors for re-ranking, e.g. warm-start, fairness etc. Built a web application by defining a RESTful-like API using Flask, built a HTTP/1.1-compliant, WSGI thread-pooled web server on a Spark cluster using CherryPy.