Large model AI safety engineer (Shenzhen)

Roles & Responsibilities

1. Responsible Tencent's AI product security evaluation work identifies potential vulnerabilities and risks in AI's endogenous security (including but not limited to algorithm model security, infrastructure security, etc.) in advance, and proposes corresponding solutions
2. On the cutting edge Track and analyze technologies in the field of AI security attack and defense, and combine them with Tencent AI business scenarios to improve attack verification capabilities and defense strategies
3. Conduct AI security research on common open source Foundation Models, and implement them in practical applications and business empowerment Explore other aspects.

Qualifications

1. Bachelor degree or above, computer science, network security or related majors
2. In LLM security (prompt injection, adversarial samples, model backdoors, poisoning attacks , model stealing, etc.), and understand LLM alignment, robust training and other security protection strategies
3. Proficient in at least one programming language, such as Python, C++, Java, etc., with strong coding skills Ability to use existing SOTA methods for training and testing, and be able to undertake the development of project sub-modules
4. Have good communication and teamwork skills, strong analytical and problem-solving abilities, and be able to quickly identify and solve security problems, and have a high interest in security attack and defense.
5. Bonus points: Priority will be given to those with relevant working experience in AI Red Team those who have published papers/articles in top conferences/journals (ACL, ICLR, NeurIPS, ICML etc.) will be preferred and those with open source project research and development, security Priority will be given to those with experience in winning competitions.