1. Built and maintain a suite of web-based tools for metabolite analysis and search by Python andJavaScript .( http://www.omicscraft.com/tools/)2. Implemented data preprocessing of the bioinformatics dataset from laboratories and users.3. Using Artificial Neural Network to metabolite identification. Compared with traditional methods, the identification accuracy is improved by 12%.

Made APP part of the person profile for Bixby (virtual assistant of the cellphone): 1. Collected the App data from three APP store by coding web crawler, and analyze the data. 2. Used classification algorithms (SVM, RNN, FastText, etc.) to classify APP and game, and analyze the performance. (The accuracy of the result is 6% higher than the Korean Bixby version) 3. Add the tags to each category, design some strategies for the recommended system.

Studied deep learning by image classification. Built the Caffe platform, and analyze the Caffe code to learn inner mechanism. Ran models of Imagenet classification, and analyze the performance. Used the Caffe to make other deep learning study.