Pet Store Web ApplicationJan 2022 - May 2022
- Proposed E/R diagram for a store management system, created SQL schema for database and implemented it with PostgreSQL.
- Designed back-end of web application based on Flask framework, improved its vulnerability through SQL injection prevention.
- Developed a user-friendly front-end with HTML and CSS. Users can view products, add/remove items to/from shopping cart, make appointment by checking clerk's timeslots. Merchants can manage inventory and update staff availability through the system.
- Deployed the website on Google Cloud Platform to make it externally accessible.
Twitter Sentiment Data AnalysisSep 2021 - Dec 2021
- Preprocessed 1.6 million Twitter text data with different embedding methods: TFIDF, Word2Vec and CountVectorizer.
- Cooperate in a group of 4 to built multiple traditional NLP models to do sentiment classification.
- The best one can analyze millions of tweets in one minute and reached 80.3% precision, which beats LSTM in both accuracy and efficiency.
The Real APPJan 2019 - May 2019
- Proposed a command line based data management system for Real Estate Agent to manage large number of clients.
- Designed functions: archive/unarchive, pin/unpin, select/unselect, redo/undo, implemented "Pin/Unpin Contact" features and resolved mixing command problems.
- Embedded Google Map in applet for various purpose, and enriched Graph User Interface.
Skin Lesion ClassificationSep 2018 - Dec 2018
- Applied random data augmentation techs including flipping, rotation and zooming to images to avoid overfitting.
- Developed convolutional neural networks with Keras to classify dermoscopic images into three types of skin diseases.
- Leveraged pre-trained ResNet as a feature extractor, apply transfer learning techniques train a customized classifier.