Jiashu Chen

• Designed and built a full-stack recommendation and customer management system for restaurants • Implemented the Restful API with Java Spring, retrieved transaction data from database and converted to RFM model • Processed the data and used different methods to segment the customers including K-means and Fuzzy C-means model • Used BG/NBD model to analyze and predict customer behavior with an accuracy of 80%, calculated customer value based on results • Stored and trained data with GCP BigQuery and used Spark to accelerate the training process, reducing the training time by 90% • Use Airflow to host the system and automatically sending reminder emails to recall lost customers based on prediction results

• Collaborated with 3 people to research the field of reinforcement learning and implemented DQN algorithm • Utilized AlphaZero to train a Gomoku agent and eventually reached the level of human control • Scraped image and text data from the Internet, preprocessed the data, and built an inverted index • Utilized ResNet50 and Word2Vec to train and classify 100,000 photos with keywords and reached an accuracy of 82% • Implemented a Photo Search Engine via Flask and the deep learning model to get corresponding image results by entering keywords