Adversarial Prompt Expert

company:Reinforce Labslocation:Remoteremote:Remotecomp:COMP NOT DISCLOSED

# editor_note · curation

Red-team role finding jailbreaks, ranking model failures, and documenting attack paths so safety teams can patch them.

source_signals.ymlwhat justified inclusion

✓adversarial_testing_focus:Design and execute complex jailbreak attempts to identify vulnerabilities in state-of-the-art models

✓systematic_llm_evaluation:Model Evaluation: Systematically rank LLM outputs to determine where safety guardrails are failing or succeeding

✓documentation_for_patch_teams:Knowledge Loop: Document your "attack vectors" clearly to help our engineering teams patch vulnerabilities

✓creative_evasion_required:Proven ability to navigate complex model restrictions using creative evasion techniques

✓no_coding_requirement:You don't give up when a model says "I cannot fulfill this request." You find a new angle

✓linguistic_social_science_background:Use your background in linguistics or social sciences to find "hidden" biases or harms

tagsarchetype + ai-native surface

#adversarial testing#LLM safety evals#jailbreak research#prompt security#red teaming

descriptionsource excerpt

Location

Remote

Employment Type

Contract

Department

Expert Operations

Compensation

$20 – $40 per hour

We are looking for a creative "breaker" to join our team as an Adversarial Prompt Expert. In this role, you won't just be using LLMs—you'll be stress-testing their boundaries, bypassing their safeguards, and helping us build safer, more robust intelligence.

This is an asynchronous, remote position designed for self-starters who thrive in the gray areas between code, linguistics, and security.

Work Details

Design and execute complex jailbreak attempts to identify vulnerabilities in state-of-the-art models.

Use your background in linguistics or social sciences to find "hidden" biases or harms that standard automated filters miss.

Model Evaluation: Systematically rank LLM outputs to determine where safety guardrails are failing or succeeding.

Knowledge Loop: Document your "attack vectors" clearly to help our engineering teams patch vulnerabilities.

Who you Are

Heavy LLM Usage — hands-on experience with multiple models (open- and closed-source), comfort experimenting across systems and platforms.

You have a "hacker mindset." You enjoy the puzzle of finding edge cases and can think of ten different ways to ask a forbidden question.

You can turn a chaotic afternoon of prompt-hacking into a clean, actionable report.

You understand the weight of this work. You can handle sensitive or "dark" content professionally and stay within ethical boundaries.

Qualifications & Skill Requirements

Proven ability to navigate complex model restrictions using creative evasion techniques.

Background in offensive security or red teaming is a major plus.

You don't give up when a model says "I cannot fulfill this request." You find a new angle.