AI Red Team Engineer for LLMs

The work

You will evaluate and red team large language models and AI agents for safety issues and security vulnerabilities. You will create repeatable tests that can run offline and be evaluated automatically.

The work also includes building scripts, custom tools, test environments, and test harnesses. You may contribute to security research, create realistic attack scenarios, advise on secure coding and platform hardening, and document findings with clear steps for reproducing them.

• Test AI models and agents for safety risks and vulnerabilities.
• Create offline, reproducible, and automatically evaluated test cases.
• Build automation scripts, custom tools, environments, and test harnesses.
• Design realistic attacks and document findings with clear reproduction steps.
• Work with LLM risks such as prompt injection and the OWASP Top 10 for LLMs.

What it pays and takes

This is an intermediate-level, part-time contractor role. You must be able to work in English at C1 level or higher and submit your resume in English.

• Pay: $40 per hour.
• Hours: Fewer than 20 hours per week with flexible scheduling.
• Work setup: Remote and open worldwide, subject to location eligibility.
• Education: Bachelor’s or Master’s degree in computer science, software engineering, cybersecurity, digital forensics, or a related field.
• Certification: At least one recognized cybersecurity or red-team certification with a public Credential.net link for verification.
• Technical skills: Strong scripting in Python, Bash, or PowerShell.
• Security experience: Penetration testing across web, API, network, and infrastructure systems.
• Tools and methods: Experience with Docker and CI/CD security tooling.
• LLM security: Familiarity with prompt injection and the OWASP Top 10 for LLMs.
• Helpful experience: Frameworks such as garak or PyRIT.

How it works

Apply on OpenTrain. The employer reviews applications there.

About AI training work

AI training work uses human expertise to test, review, and improve artificial intelligence systems, including checking whether models respond safely to realistic attacks. OpenTrain AI is the hiring and contracting organization for this role, and experienced cybersecurity contributors help make these systems more secure.