Tech led a high-visibility client front-facing reversed proxy migration project from monolith to distributedmicroservices. Designed authentication exchange with API gateway, session sharing using Redis to improvescalability, docker image, and Helm chart for service deployment.Designed on-premises bare metal Kubernetes, infrastructure services, and back-end microservices for an analytic product target to upgrade 50% of global clients. Infra included Timescale, HDFS, Trino, and Kafka with HA and scalability. With back-end services like ingestion handler and db keeper. Onboard junior developers to the project.Worked with Scalla scaffolding data ingestion handler as a Kafka consumer. Allowed faster data processing with maintained in-memory metadata mapping from API or offline files.Optimized database storage and query performance with customed micro-service. Utilized chunks and compression into columnar format. Reduced storage capacity by over 50% and increased read performance from second to ms.Implemented middleware proxy service to firewall connection interfaces in gRPC, custom HTTP, gRPC streamed HTTP. Developed file cache increased web service response performance by over 20%.Patched security bugs related to XSS, code injection, and sensitive token leaks. Designed a token exchange process that cooperates with the front end and gateway to align with best security practices.

Designed and developed Computer Vision Training Service. The service takes photo and metadata and automatically trains recognition model, where user can later recognize object through streaming video to the service.Experimented two machine learning networks for object recognition in TensorFlow. Researched for the most suitable network between training cost, recognition efficiency and accuracy that can be used as video recognition.Integrate trained model with live streaming service. Allows the user to recognize trained objects through web camera. Supports both mobile and web.Collaborate with team using Club House, an Asana-like team management tool, and integrate through GitHub. Weekly meeting in New York Verizon office with senior designer and manager, and a separate team standup meeting.

Trained machine learning model using miniSOM to classify one year of outlier data collected from firewall across the globe. Improved the accuracy of choosing interest outlier compared to third deviation classification.Developed full-stack graph generation API service using Python, Flask, Nginx, uWSGI, fastcgi on Windows and Ubuntu server for both QA and production; integrated into Qlik view to support enterprise planning and decision making.Designed modularized structure coding using blueprint in Flask to isolated module from web server. Enabled future development to be implemented without interfering with server configuration. Maximized the capability to develop or experiment with customized functions. Added server status tracker for different modularized code that was more efficient.

Designed and programmed configuration/migration program for BIG-IP F5 using Python. Moved up project schedule two-months earlier compared to manual configuration migration; the project is later promoted into production and serves as one of the product-service features.Managed and planned load balancing and WAF for two major banks’ public domain web service.Tutored and trained 4 new interns in network and programming areas to ensure a smoother onboarding process and lay the foundation for a better learning experience.Reviewed and tested reported CVE threats against simulated client network environment. Provide method of temporary blocking and long-term update plan or advice.

Computer center general IT support. Responsible for maintenance of all IT services within the computer center, which includes repair malfunction computers, network structure modification, TA for computer center provided class(PC DIY), maintain schools IT services such as printers or authorized software and school IT infrastructure on phone troubleshoot.

Design medical Web application and server coordinate with a smart blood pressure monitor. The application enables measure data to be transferred in private server and later be confirm by profession medic personals on cellphone or desktop web page before storing into HIS. The confirmation process is real-time synchronized and display on all authorized client side. Which provides a more efficient method of managing collected data. The application was complete with C#.Net and Web Socket.