微信安全-大模型安全高级工程师
Core
Design and implement a closed-loop security system for large language models (LLMs) covering risk identification, assessment, mitigation, and iteration, while conducting research on adversarial attacks and defense mechanisms.
Role type
Senior IC LLM Security Engineer
Builds
Automated red team evaluation systems, blue team defense frameworks, and security assessment tools for LLMs and AI agents.
Domain
Artificial Intelligence / Cybersecurity
Deliverable
production ML models
Required skills
LLM architecture design, prompt injection defense, model alignment, red team testing, adversarial sample detection, model watermarking, vector database security, API security, data-driven strategy optimization, SFT/DPO/RAG/Agent implementation
Preferred skills
Experience with generative AI compliance policies, privacy protection, and global security landscape analysis
Technologies
SFT, DPO, RAG, Agent, Vector Databases, AI APIs
Responsibilities
Lead the planning and construction of the LLM content security risk control system; Research and implement advanced security mechanisms against hallucinations, data poisoning, and privacy leaks; Build automated red/blue team capabilities and regression testing mechanisms; Optimize security strategies using data-driven analysis; Design security architectures for vector databases, agents, and AI interfaces; Monitor global compliance trends and translate them into product security practices.
