Responsible for AI Agent security review and penetration testing, including defense design and implementation against emerging threats such as AI tool misuse, context poisoning, various forms of data/prompt poisoning, and prompt injection.
Responsible for designing and developing AI-based automated security detection tools, enabling automated alert triage, attack attribution, and improved security operations efficiency.
Participate in building Agent behavior auditing and anomaly detection systems, identifying abnormal behavior chains during Agent runtime, constructing an AI-driven security operations platform, and exploring, designing, and implementing defense and countermeasure solutions based on real business scenarios.
Combine security risks of cloud-native infrastructure (K8S, Docker, cloud services) to assess the attack surface of Agent runtime environments and design defense-in-depth solutions.
Requirements:
At least 3+ years of security offense/defense experience (penetration testing/red team practice), with 1+ years of experience in LLM (large model) security offense/defense; extensive hands-on red team experience preferred.
Familiar with network architecture and cloud-native technology stacks, including K8S, Docker container security, and cloud security (IAM, network isolation, cloud-native threat models, etc.), with the ability to assess security risks in complex infrastructure.
Possess practical Agent development capability, able to independently build/modify AI Agents (e.g., based on frameworks such as LangChain), as well as hands-on offensive capability against Agents (practical construction and validation of attack techniques such as tool misuse, prompt injection, context poisoning).
Familiar with security frameworks such as OWASP Top 10 for LLM and MITRE ATT&CK, familiar with common AI attack techniques, adversarial logic, and defense solutions, with a strong passion for AI security.
Proficient in at least one programming language such as Python/Go/C++, with the ability to develop security offense/defense tools or perform deep secondary development on open-source security platforms.
Nice to Have
Familiar with mainstream AI frameworks (PyTorch, TensorFlow, LangChain, etc.), with practical experience in local deployment and optimization of large language models.
Hands-on cloud security (AWS) experience or related certifications.
