AutoAudit is an LLM-powered intelligent audit platform that integrates knowledge graphs, retrieval-augmented generation (RAG), and reinforcement learning to support audit decision-making. It provides a multi-turn dialogue agent based on LangChain for professional reasoning and audit suggestions, vector database-backed semantic search, and supports model training with SFT, RLHF, and LoRA fine-tuning. The system covers risk assessment, compliance checking against standards like COBIT and ISO27001, and continuous model improvement through benchmarking.