Utilize LLM to recommend AML rule improvement plan based on True Positive / False Positive pattern and automatically generate AML rules on demand.

Utilized Databricks and PySpark to aggregate complex datasets into high-level performance rule metrics including but not limited to True Positive rate and F1 score, and visualized the results.

Maintain existing Python monolith to ensure efficient rule execution and data processing and continuously improve the feature set for the customers.

• Developed and deployed a full stack application using Next.js, MongoDB, and AWS Elastic Kubernetes Cluster, incorporating a Network Load Balancer (NLB) and automating cluster management with MongoDB Operator and AWS Load Balancer Controller to allow automatic horizontal scaling.

• Implemented cert-manager to handle SSL certificates across the cluster, ensuring secure and reliable application performance.

• Built a real-time chat server in Rust with Socket.IO which supporting automatic resync to handle disconnection events.

• Built a spatiotemporal model with ArcGIS and Transformer model utilizing 110+ years wildfire data & 200GB+ of Landsat satellite data, including location, spread, duration etc., applied statistical analysis, anomaly detection, outlier analysis, time-series analysis to understand the seasonal behavior pattern of California wildfires.

• Deployed the model into prediction to predict the future trend of wildfire with multiple factors including past wildfires, precipitation, and vegetation index, and achieved 70% accuracy (Python).

Manage Uni, the internal scheduling and management system at Unitrans, to make sure it works as drivers and managers rely on the system to check their schedules and status of vehicles. Also manage the Unitrans public website.

Manage the Virtual Machines belong to Creative Media, develop new web softwares with Django to digitalize the workflow at Creative Media, and operate the Kubernetes cluster of Creative Media.