AI Red Team Engineer for LLM Security

The work

You will design and run adversarial security tests against large language models, AI agents, and retrieval-augmented generation pipelines. The work focuses on finding weaknesses, measuring model behavior, and documenting clear ways to reduce risk.

You will work remotely on a part-time contract, follow detailed project guidelines, and uphold strict ethical and safety standards.

• Create and automate attack prompts, including prompt injection and jailbreak attempts.
• Build test suites for LLMs, agents, RAG pipelines, function calling, and tool use.
• Define scoring rubrics and grade how models respond to adversarial inputs.
• Document reproducible findings with risk ratings and propose practical mitigations.
• Contribute small Python, Bash, or PowerShell scripts and utilities to scale testing.
• Assess attack surfaces involving data exfiltration, web systems, APIs, networks, and infrastructure.

What it pays and takes

This is an intermediate-level contractor role for people with hands-on cybersecurity, penetration testing, and LLM security experience.

• Pay: $40 per hour in USD.
• Hours: Less than 20 hours per week.
• Work setup: Remote and available worldwide.
• Work type: Part-time contract.
• Language: Advanced English proficiency at C1 level.
• Experience with Python, Bash, or PowerShell.
• Experience with containerization and CI/CD security, such as Docker.
• Experience testing web applications, APIs, networks, and infrastructure.
• Knowledge of LLM attack surfaces, including prompt injection, jailbreaks, and data exfiltration.
• Knowledge of the OWASP Top 10 for LLMs.
• Ability to use or red-team with tools such as garak or PyRIT.
• Availability to complete a HackerRank and platform assessment as soon as possible.

How it works

Apply on OpenTrain. The employer reviews applications there.

About AI training work

AI training is the human work used to improve artificial intelligence, including testing model responses, finding unsafe behavior, and rating results against clear guidelines. OpenTrain AI hires and contracts contributors for this work, and experienced specialists are needed when projects require deep security and technical knowledge.