Responsibilities include:

• Research on the state-of-the-art model for Visual Question Answering Problem
• Built the end-to-end pipeline based on Kubeflow from data collection, model training, testing to deployment
• Deployed the model on KFServing, which is based on KNative and provide auto-scaling
• Developed an Web application for Demo, using React as front end, Django as backend and PostgresSQL as database, which shows the idea that the serving data can also be collected and used for training after reviewing
• Leader: Ed Henary, Distinguished Member of Technical Staff

Responsibilities include:

• Done research on the Heterogeneous Face Recognition utilizing adversarial learning and correlation analysis
• Wrote automatic tools for efficient traning data prepration and easy-to-configure data loaders
• Repruduced latest algorithms in ICCV/CVPR papers
• Advicer: Dr. Zhifeng Li, Principal Researcher and Dr. Dihong GONG, Senior Researcher

Joining the Lab as a CSST (the Cross-disciplinary Scholars in Science and Technology) program member, my responsibilities include:

• Proposed a new neural network structure which introduced the concept of a segmentation template to better segment the brain structure of 3D MRI compared to some state-of-the-art neural networks such as 3D-UNet and DeepMedical
• Our model outperformed existing models with respect to the model size and the prediction accuracy, and thereby was better capable of assisting doctors when they are diagnosing brain images
• Advicer: Dr. Lei He, Professor