Project – Mobile/Multi-access Edge Computing (MEC) Middlebox
• Constructed multi-cluster orchestration platform in MEC middlebox integrated with Kubernetes.
• Utilized OpenAirInterface, radio access network, LISP tunneling, multi-cluster orchestration system.
• Supported 23 DNS providers for end-users to customize domain name of services.

Product – Kyronus
Developed cross-platform mobile simulation game championing Taiwan culture of historical site.
Key tech: mobile application framework, serverless computing, cloud, software engineering, DevOps.
• Proposed and integrated Apache Cordova into project to build one-codebase hybrid (Android & iOS) mobile app.
• Shortened app development cycle by proposing and integrating Apache Cordova into project to build an one-codebase hybrid mobile app (Android & iOS).
• Avoided generating excessive load to API server by implementing truncated exponential backoff on front-end API requesting in Axios interceptors.
• Improved frame rate up to 200% in map-related operation (scaling, rotating, relocating, ...etc) by building Cordova native plugin with Mapbox SDK.
• Introduced DevOps pipelines using GitLab CI/CD and Docker for Android automatic deployment in semantic versioning and thus successfully shortened development cycle time by up to 50%.
• Reduced Docker image's size by up to 300% with Multi-Stage Builds tech to accelerate the build and deployment processes.
• Led 7-person team and coordinated all Scrum Ceremonies including Sprint Planning, Daily Stand-ups, Sprint retrospectives, Sprint Demos, Story Grooming, and Release Planning resulting in on-time delivery rate of 95%.
• Built multiple kernel components of game mechanisms: collect resource system, mining system, offline countdown timer, and so on.
• Constructed several back-end APIs for collect resource system, email system using Express.js, DynamoDB, AWS Lambda, and AWS Gateway.
…
see more

Project – KPI dashboard
Constructed web-based dashboard to produce detailed monthly KPI reports automatically.
Key tech: web UI library, database, Object-relational mapping, PSD to HTML, secure transport protocol.
• Used jQuery to built mobile UI adopting Responsive Web Design to work smoothly on both mobile and desktop devices.
• Build real-time dashboard chart with Apache ECharts and implemented periodically API requesting using JavaScript closure.
• Enabled HR to import KPI data from Microsoft Excel spreadsheets into Django model.
• Automatically generated monthly KPI reports using shell script and Linux crontab and thus decreased HR daily workload by 13%.
• Directed 3-person team and completed the project 2 weeks in advance with an additional 13-page documentation slashing job handover.
• Proposed AAA (Authentication, Authorization, and Accounting) secure transport protocol using Kerberos and Microsoft Active Directory.

Project – Fraud call detection
Trained a ML model for fraud call detection.
Key tech: classifier, feature engineering, data pre-processing.
• Utilized Recursive Feature Elimination (RFE) and bagged decision trees to reduce number of features from 52 to 20 and successfully reduced training time.
• Applied weighted average ensemble (LightGBM, GradientBoosting, XGBoost, CatBoost, and RandomForest) to final prediction and thus improved precision rate by up to 10%.
• Built a multi-classifier with 94% precision to detect fraud call with FarEasTone Telecom's anonymous data.