Distributed Systems

● Lung Nodule product: Built 3D Mask R-CNN based model for lung segmentation and trained 3D U-Net based multiple head model for lung lobe segmentation by PyTorch. Added as a new feature in a new released version to assist in lung lobe 3D visualization and lung nodule localization.

● DevOps: Initialized portable R&D environment by Docker images and container registry service on Alibaba Cloud across several research teams. Maintained Jenkins automation test tools for CI/CD pipeline and conducted release tests for several products. In collaboration with customer service team to optimize product feature based on customer feedback.

● OCR product: Created a back-end service and implemented related API to collect medical document image from different third-party medical devices and recognize patient ID via Optical Character Recognition (OCR) technique. Trained U-Net based deep learning model by Python and TensorFlow on synthetic text images with data augmentation. Obtained 96% accuracy on character recognition and outperformed previous version by 5% for large-scale release test.