Pittsburgh, Pennsylvania, United States
• MCP Server: Designed and implemented a Model Context Protocol (MCP) server on AWS Lambda, enabling AI agents to securely access internal Coral HTTP services. • Authentication and Throttling: Built end-to-end infrastructure with AWS CDK, integrated with API Gateway, IAM authentication, and CloudAuth for cross-account access. Developed custom SigV4-compatible MCP adapters and implemented dynamic throttling (SDC) for scalable, secure request handling. • AI Agent Support: Created a Python-based Coral client library, resolving dependency conflicts, handling custom data types, and integrating with LangChain agents. • Deliverable: A reusable helper library to simplify future MCP deployments, improving standardization and reducing onboarding time for other teams. Wrote design documentation and conducted internal design reviews, aligning technical decisions with team requirements and best practices.
Background developer using Java. It was a web application for Pipechina’s online trades. Used Spring boot framework and dubbo service. Evolved several design patterns like faéde and service, mapper and controller. Designed the whole processing logic and wrote relative document for a specific function. Shared the ideas of this design and assigned work. Finally finished the whole tech stack for the borrow and return gas function.
Designed an Android application named MWalk, which can demonstrate daily and weekly step ranking and realize real-time interaction between users Responsible for implementing the login interface and retrieving the database interface. The whole UI of this app was designed in Android Studio, using Java. Raised the idea of community and competition. Then developed the competition backend in Firebase. Also searching into Firebase safety protocol to solve users’ access right problem. MWalk had about 30 users after the exhibition. Almost all of them highly appreciated it.