AI Safety Red Teaming Specialist

About OpenTrain

OpenTrain AI is the centralized platform for people who build their careers in AI training and data labeling. We help freelancers discover specialized AI training projects, consolidate proof-of-work into a single profile, and grow a durable portfolio that showcases real adversarial and evaluation experience.

OpenTrain AI is the hiring and contracting organization for this role. We connect experienced contributors to high-impact safety work that helps shape how modern AI systems behave.

Why AI Training and Red Teaming Matters

AI training (data labeling, annotation, and human feedback) is the human side of modern AI: people create, evaluate, and refine the examples and tests that models learn from. Red teaming is a specialized subset that stress-tests models with adversarial prompts, jailbreaks, and multi-turn manipulation to find real-world failures before they cause harm.

This work is remote, flexible, and directly influential—contributors help make models safer, reduce misuse, and improve trustworthiness across products and research.

The Role

OpenTrain AI is recruiting an AI Safety Red Teaming Specialist to stress-test conversational AI systems using adversarial prompts and multi-turn attacks. You will review text outputs, classify failures, and produce reproducible artifacts (reports, datasets, and attack cases) that teams can act on.

This is a contractor, part-time role for experienced red teamers who can apply structured frameworks and clearly document vulnerabilities for both technical and non-technical stakeholders.

What You'll Do

Your day-to-day will focus on designing and executing adversarial tests, annotating failures, and producing clear deliverables that capture reproducible attacks and systemic issues.

• Red team conversational models and agents with jailbreaks, prompt injections, misuse cases, bias exploitation, and multi-turn manipulation.
• Annotate failures and classify vulnerabilities using project taxonomies and benchmarks.
• Follow playbooks and structured frameworks to keep testing consistent and reproducible.
• Document findings in reports, datasets, and attack cases that downstream teams can act on.
• Communicate risks and remediation priorities to both technical and non-technical stakeholders.

Requirements

To succeed in this role you must bring direct adversarial testing experience and strong communication skills. OpenTrain AI requires native fluency in both English and Bengali for this posting.

• Prior red teaming or adversarial evaluation experience in AI, cybersecurity, or socio-technical probing.
• Familiarity with jailbreaks, prompt injection, multi-turn manipulation, and ability to classify vulnerabilities.
• Comfort using structured frameworks, taxonomies, or benchmarks instead of ad hoc testing.
• Strong written communication for producing reproducible reports and datasets.
• Native fluency in English and Bengali (required).

Helpful Background

The following backgrounds are beneficial but not strictly required if you can demonstrate equivalent red teaming experience and reproducible deliverables.

• Adversarial ML experience: jailbreak datasets, prompt injection, RLHF/DPO attacks, or model extraction.
• Cybersecurity skills: penetration testing, exploit development, or reverse engineering.
• Socio-technical risk experience: harassment, disinformation, or abuse analysis.
• Creative adversarial thinking from psychology, acting, writing, or related fields.

Schedule, Pay & Logistics

This is remote contractor work. Pay is hourly and ranges from $20 to $22 per hour. The role is listed as part-time/contractor—typical time expectations are 20+ hours per week, and this project's default commitment is stated as 40 hrs/week.

OpenTrain AI will expect reproducible deliverables (annotated datasets, attack cases, and written reports) as part of the engagement.

• Employment type: Contractor, Part-time.
• Time requirement: 20+ hours/week (default commitment: 40 hrs/week).
• Hourly pay: $20–$22 USD per hour.
• Data type: Text; label types include RLHF, evaluation rating, and red teaming.

How It Works / Apply

If this role fits your experience, apply through your OpenTrain AI profile. OpenTrain is where AI training professionals build a unified portfolio of work, making it easier to be considered for specialized safety projects.

When you apply, be ready to show prior red teaming artifacts or examples of reproducible attack cases and to confirm native fluency in English and Bengali.

• Prepare examples: red teaming reports, jailbreak testcases, or annotated evaluation datasets.
• OpenTrain AI is the hiring and contracting organization for this role; submit work history and deliverables via your OpenTrain profile.