• Worked on Huawei Cloud AI and algorithm team for short-term prediction of the expansion of physical servers.• Trained a Prophet model to forecast 13-week user requests on virtual machines based on past 26 weeks; implemented an LSTM model implemented by PyTorch and compared performance with the Prophet model.• Explored user’s behavior by linear modeling; input user’s behavior as useful features to the Prophet model and decreased RMSE by 5%.

• Developed a recommendation system model with latent factors to make predictions related to user/short-form user video interactions.• Extracted useful features through sentiment analysis, and improved the prediction accuracy by more than 8%.

• Implemented automatic metrics to evaluate a new interactive language generation framework that allows users to read and control the mind of GPT2.• Analyzed results of evaluations and showed that the new interactive language generation framework we presented outperformed several variations of classic clustering methods, and the Plug-and-Play Language Model from Uber AI.

• Designed and developed a digital platform to fetch and calculate metrics data from several machine learning training record tables by NodeJS, hosted the service on company’s server.• Displayed data on web pages in the form of line charts and pie charts by ReactJS; interacted with the database using MySQL.