Algorithm Engineer · World Model · AI Agent · Team Lead
📍 Shenzhen, China
📧 Email: lintsinghua@qq.com
Open to technical consulting and collaboration.
AI algorithm engineer and open-source developer with 11k+ GitHub Stars across personal open-source projects, specializing in AI Agents, World Models, and Model Inference Acceleration.
Original creator of DeepAudit, a multi-agent security auditing platform, and author of Claude Code Book, an in-depth exploration of Agent Harness architecture. Contributor to HelixWorld’s real-time audiovisual world model research and YuE2-Turbo’s inference acceleration for music generation, with hands-on experience spanning algorithm research, system optimization, and production deployment.
Open-source multi-agent system for code vulnerability discovery
An AI hacking team for the public, making vulnerability discovery accessible to everyone. As China's first open-source multi-agent system for code vulnerability discovery, it features one-click deployment for beginners, autonomous collaborative auditing capabilities, and an automated sandbox vulnerability verification (PoC) mechanism. It supports Ollama private deployment and one-click generation of audit reports, democratizing security technology and simplifying code auditing.
A 420K-word deep architectural analysis of Claude Code and Agent Harness design
A systematic book on decoding Agent Harness architecture through Claude Code, covering conversation loops, tool systems, permission pipelines, context compression, memory, hooks, sub-agent orchestration, MCP integration, skills, streaming architecture, Plan mode, and the practical path to building your own Agent Harness.