Wanwei Huang Email & Phone Number

Redmond, Washington, United States

Shenzhen, Guangdong, China

Enhanced the accuracy of the Decision Tree model for anomaly detection in automated friend requests to prevent freezing of robot accounts. Addressed the issue of low accuracy and explored alternative detection methods, introduced new features, and ultimately achieved a 10% decrease in the number of robot accounts being frozen Created docker and conducted performance tests on manually designed prompts for generating dialogues using large pretrained language models (PLMs), and explored… Show more Enhanced the accuracy of the Decision Tree model for anomaly detection in automated friend requests to prevent freezing of robot accounts. Addressed the issue of low accuracy and explored alternative detection methods, introduced new features, and ultimately achieved a 10% decrease in the number of robot accounts being frozen Created docker and conducted performance tests on manually designed prompts for generating dialogues using large pretrained language models (PLMs), and explored transfer learning to develop a Chinese language modelImplemented a reinforcement learning model based on a research paper and devised case-specific scores and rewards to refine the prompt generation for dialogue generation, preserved the knowledge from the original large PLM Finetuned story-generating PLM by training it on Project Gutenberg data and automated generation of storiesAutomated workflow to parse 5+ Named Entity Recognition datasets and conducted experiments on new model Show less

Shenzhen, Guangdong, China

Assessed performance of 10k+ companies over 2020, extracted primary statistical features from 140+ features via PCATrained Random Forest Classification model on data from 2019, then retrieved and tokenized natural language data from database, resulting in a 10% increase in accuracy Implemented Grid Search Cross-Validation, improved the performance of the model and tackled the dataset's class imbalance by down-sampling, resulting in a remarkable 25% increase in accuracy Visualized… Show more Assessed performance of 10k+ companies over 2020, extracted primary statistical features from 140+ features via PCATrained Random Forest Classification model on data from 2019, then retrieved and tokenized natural language data from database, resulting in a 10% increase in accuracy Implemented Grid Search Cross-Validation, improved the performance of the model and tackled the dataset's class imbalance by down-sampling, resulting in a remarkable 25% increase in accuracy Visualized the performance of the model with AUC-ROC curve and showcased the importance of the newly-obtained data. Proposed regulation of linguistic input from users and ensured the accuracy in the long run Show less

Shenzhen, Guangdong, China

Self-taught Vue.js and the components came along with the frame, including XML, JavaScript, HTML and CSS. Created internal webpage that allowed managers to upload test scores of employees, to extract data about scores from MySQL database using XML, to filter on the webpage using JavaScript and to export the results on the webpage to Excel files.

Master Of Science - Ms, Data Science

Bachelor Of Arts - Ba, Mathematics