At Can Tho University Software Center (CUSC), I developed an agentic Retrieval-Augmented Generation system using LangChain Tool Calling Agents for dynamic tool execution. I also implemented session-based retrieval with Redis caching and integrated Cohere Rerank.
I evaluated the pipeline with RAGAS, achieving an overall score above 0.92. The system supported knowledge-base and file-based question answering.
On LexMind - Vietnam Law Chatbot, I built a multi-agent RAG system for Vietnamese legal document Q&A, with a self-correcting reasoning pipeline and grounded legal citations. LangGraph Supervisor orchestrated four specialized agents.
I also developed a RAG-based virtual assistant for Vietnamese legal PDF questions, combining semantic search with BM25 and generating answers with grounded citations. I’m interested in scalable LLM applications, AI knowledge systems, and real-world AI integration.
