Résumé | HU, CHENKAI

Education

Building audio-domain Agent systems, focusing on speech synthesis, voice cloning, and audio effect generation pipelines.
Exploring Agent-based architectures for audio understanding and generation tasks using multimodal LLMs (e.g., Qwen3-Omni).

Built a Multi-Agent collaborative workflow using LangGraph with Chain-of-Thought task planning and state machine scheduling, automating requirements document generation and iteration — improving writing efficiency by 90%+.
Designed a Tool Calling mechanism and standardized MCP Server wrappers (document analysis, text validation) based on the Model Context Protocol, decoupling model inference logic from external tools and databases.

Built a Retrieval-Augmented Generation (RAG) system for an AI companion app; applied embedding-based indexing and retrieval over conversation history to construct dynamic context, significantly improving long-context memory accuracy and coherence.
Designed scalable backend services using RESTful architecture to support stable, production-level dialogue services.