1. 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.
2. 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.
3. 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).
4. 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.
5. 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

1. Familiar with mainstream AI frameworks (PyTorch, TensorFlow, LangChain, etc.), with practical experience in local deployment and optimization of large language models.
2. Hands-on cloud security (AWS) experience or related certifications.