· Implemented an automation system that can reclaim and get status of the unused Access Point devices. The automation system replaced the old manual and insecure system. This new tool greatly simplified the AP device management workflow, and largely improved the teams' work efficiency.· Implemented an internal tool from end to end using JAVA, Ruoyi framework, Mybatis, Shiro, Thymeleaf to enable the teams to retrieve relative backend data such as clients info, Access Point device info, device network policy update info in nearly real-time for business usage and data insight which greatly enhanced the work efficiency on both sides. · Implemented and expanded external API services including several different aspects of endpoints such as logs, devices and clients on top of the existing XAPI framework using gRPC, JAVA and Spring framework. · Designed and improved the architecture of the ETL data pipelines using AWS services like S3, Sqs, Lambda function, Glue, Athena, and fully automated the deployment process of the ETL pipeline using python and shell scripts to AWS cloud platform with the secure access control enabled.· Developed a global metrics dashboard that collects important representative data in production system by using AWS firehose stream service, python and EKS to achieve data collection, delivery, visualization and monitoring. It provides a good way to display our cloud service work load and visualize how big scale the system are to the customers.· Implemented a Jenkins pipeline for QA teams to automatically deploy/upgrade the test environment data center in the cloud, and automatically deallocate the system, which greatly increases the independence of the deployment/upgrade process for other teams.· Lead a four people team to complete the system architecture migration from the Instance-Level data center to Pod-Level data center using AWS Kubernetes service, Helm, Terraform and etc.

· Working as a data analyst in BiLAB of NYU urban and civil engineering department· Used collected data to train the model to see how different filtered factors effect AHU power consumption based on python.· Implemented preprocessing(correlation, removing outliers, resampling), Boosting Tree and Random Forest to filter factors among hundreds of collected features and used k-means clustering to get different motifs.