Implemented an anomaly detection pipeline for complex telecommunication networks. Increased automation and reduced operating expense for telecom companies. Visualized the the results for engineers to determine the root causes of anomalies faster. Published the models, usages and outputs with Notebook so data scientists and consultants at Bell Labs can easily understand the method. Analyzed the values and constraints of the pipeline and estimated its business value.

Implemented machine learning models to reduce the number of false alarms generated by the company's lost prevention IoT device. Applied data pre-processing and exploratory data analysis to identify useful features in various use cases. Cleaned erroneous data generated by incorrect software logic. Designed and implemented a multi-layer clustering model to predict safe locations for users. Visualized the result of prediction on Google Maps to assess performance using known safe locations. Help created a classification model to predict if a lost alarm is valid based on data captured by smart phone. Designed a predictive analytics plan for the future through personal prediction, collaborative prediction, and prediction with third-party data.

Supervised by Prof. Yaqian Zhou at the Institute of Media Computing and Web Intelligence. Developed a mobile App with text-to-speech features for school children to complete dictation homework. Developed backend APIs and frontend websites with PHP and JavaScript to let teachers efficiently upload and retrieve course materials and monitor student performances. Developed an automatic evaluation system using OpenCV to assess children's handwriting skills through uploaded photos. Analyzed students' study habits. Designed and implemented a multimodal machine learning pipeline to segment hours-long course recordings into single-topic sections, utilizing text and video, data. Implemented and evaluated various supervised classification models including SVM, CNN, LSTM, and fastText.

Developed a customized Horizon dashboard for OpenStack to support user self-registration. Developed backend APIs with Django to accept user payment and allocate computation resources.