AI Red Team Engineer — LLM Security & Pentesting

About OpenTrain

OpenTrain is the #1 platform for finding and building careers in AI training and data labeling. We connect skilled contributors with meaningful projects that shape how modern AI systems behave.

Work through OpenTrain to join cutting-edge AI training projects, build a profile of your security and red-teaming work, and apply quickly — all remote and flexible.

About AI Training And LLM Red Teaming

AI training (data labeling, annotation, and human feedback) is the human side of building intelligent systems. Red teaming LLMs and retrieval-augmented systems is a specialized part of that work: you simulate real attacks, find vulnerabilities, and help make models safer.

This project focuses on adversarial evaluation of LLMs, agents, and RAG pipelines — a high-impact way to apply cybersecurity skills to shape production AI behavior and defenses.

The Role

OpenTrain AI is hiring multiple AI Red Team Engineers on a remote, part-time contract basis (less than 20 hours/week) at $40 USD per hour. You will design and execute adversarial evaluations of LLMs, agents, and RAG pipelines and produce reproducible findings and mitigations.

You must be able to start the screening process immediately and complete a HackerRank plus platform assessment after initial screening. Clear advanced (C1) English is required for concise reporting and collaboration.

• Employment type: Contractor, Part-time (less than 20 hours/week).
• Pay: $40 USD per hour.
• Data type: Text. Labeling focus: Red teaming.
• Language: Advanced (C1) English required.
• Assessment: HackerRank + platform test required immediately after screening.

What You'll Do

Use offensive-security and red-team techniques to probe model and pipeline behavior, produce reproducible evidence, and help teams prioritize fixes.

Work independently on detailed test plans while collaborating with remote reviewers to ensure assessments follow strict ethical and safety standards.

• Design and execute adversarial evaluations of LLMs, agents, and RAG pipelines.
• Craft, iterate, and automate attack prompts and adversarial input generators.
• Build test suites that probe function-calling, tool use, and multi-step agent flows.
• Define scoring rubrics, grade model behaviors, and produce concise, reproducible reports with risk ratings.
• Propose practical mitigations and secure-coding recommendations based on findings.
• Contribute small scripts and utilities to scale testing and automate validations.
• Follow project guidelines, document reproducible steps, and uphold ethical/safety standards.

Requirements (Must Have)

You must meet all core technical and logistical requirements below. We will verify skills through screening and the required HackerRank + platform assessment.

This project requires demonstrable hands-on experience in both traditional penetration testing and LLM-focused red-teaming.

• Bachelor’s or Master’s in Computer Science, Software Engineering, Cybersecurity, Digital Forensics, or related field.
• Hands-on penetration testing across web, API, network, and infrastructure; familiarity with cloud and container security.
• Strong scripting and automation experience using Python, Bash, or PowerShell.
• Experience with containerization and CI/CD security tools (for example Docker); familiarity with secure SDLC practices.
• Practical knowledge of LLM vulnerabilities including prompt injection, jailbreaks, and data exfiltration.
• Familiarity with OWASP Top 10 for LLMs and offensive LLM testing principles.
• Experience with AI red-teaming or evaluation frameworks such as garak or PyRIT, and evaluating RAG pipelines.
• Offensive exploitation and reverse engineering experience (for example Ghidra or equivalent); OS security skills (Linux privesc, Windows internals).
• Ability to write clear scoring rubrics, adversarial prompts, and concise reproducible reports.
• Availability to complete a HackerRank plus platform assessment immediately after screening.
• Advanced (C1) English for clear written and spoken communication.

Nice To Have

The following are valuable but not required. If you have them, highlight relevant projects or prior roles in your application.

• Prior experience at leading AI or security organizations or previous work shipping LLM security tooling or evaluations.
• Experience scaling automated test infrastructure for model evaluations or CI/CD-integrated security checks.

Location, Eligibility, And How To Apply

This project is worldwide but restricted in a number of countries, regions, and U.S. states (listed below). Please confirm eligibility before applying. After screening, you'll be asked to complete the required HackerRank + platform assessment as soon as possible.

To apply, submit your OpenTrain profile with examples of past pentesting or LLM security work, links to code or write-ups if available, and confirmation you can complete the HackerRank assessment quickly.

• Ineligible countries: Iran, Cuba, North Korea, Syria, Sudan, Venezuela, Myanmar.
• Ineligible countries continued: China, Taiwan, Kenya, Armenia, Israel, Kazakhstan, United Arab Emirates, Netherlands, Serbia, Kyrgyzstan, Turkey, Uzbekistan, Belarus, Russia, Ukraine, Abkhazia, South Ossetia.
• Ineligible country: Switzerland.
• United States — restricted states: Alaska, Arkansas, California, Connecticut, Delaware, Georgia, Hawaii, Illinois, Indiana, Kansas, Louisiana, Maine, Maryland, Massachusetts, Nebraska, Nevada, New Hampshire, New Jersey, New Mexico, Ohio, Oregon, Tennessee, Utah, Vermont, Washington, West Virginia.
• Ineligible territories/regions (part 1): Antarctica, Aruba, Åland Islands, Saint Barthélemy, Bonaire, Sint Eustatius and Saba, Bouvet Island, Cocos (Keeling) Islands, Democratic Republic of the Congo, Cook Islands, Christmas Island.
• Ineligible territories/regions (part 2): Western Sahara, Falkland Islands (Malvinas), French Guiana, Guadeloupe, South Georgia and the South Sandwich Islands, Heard Island and McDonald Islands, British Indian Ocean Territory, Northern Mariana Islands, Martinique.
• Ineligible territories/regions (part 3): New Caledonia, Norfolk Island, Niue, French Polynesia, Saint Pierre and Miquelon, Pitcairn, Réunion, Saint Helena, Ascension and Tristan da Cunha, Svalbard and Jan Mayen.
• Ineligible territories/regions (part 4): Sint Maarten (Dutch part), French Southern Territories, Tokelau, United States Minor Outlying Islands, Holy See, Virgin Islands (British), Wallis and Futuna, Mayotte.