LLM Generalist Red Teamer — Stress-Testing & Safeguards

Apply; Seattle, Northern (WA, KY); Hybrid; USD 120,000 - 180,000; Full time

Job summary

Handshake AI is seeking an AI Red Teamer to stress-test large language models by crafting adversarial prompts and evaluating guardrails. You will work across content safety, cybersecurity, and high-risk domains, collaborating with researchers to strengthen defenses.

The role involves regular exposure to potentially disturbing content and requires strong documentation, written communication, and ethical judgment. Seattle-based, contract, 40 hours/week.

Qualifications

• Experience stress-testing multiple LLMs and evaluating guardrails.
• Ability to design adversarial prompts and jailbreak tactics.
• Ability to document experiments clearly and consistently.

Responsibilities

• Craft adversarial prompts to stress-test AI guardrails.
• Discover jailbreak and prompt-injection techniques.
• Probe edge cases to provoke disallowed outputs.
• Evaluate outputs using harm taxonomies and rubrics.
• Document experiments, including methods and results.
• Review prompts from teammates.
• Contribute to harm taxonomy development and calibration.
• Collaborate with engineers, data scientists, researchers.
• Work with potentially disturbing content; stay professional.
• Stay current on jailbreaks and model behaviors.

Skills

LLMs experience

Adversarial prompts

Content safety

Ethical judgment

Written communication

Documentation

Collaboration

Python basics

Tools

LLM APIs

Evaluation tooling

Python scripting

Job description

Handshake AI is seeking an AI Red Teamer to stress-test large language models by crafting adversarial prompts and evaluating guardrails. You will work across content safety, cybersecurity, and high-risk domains, collaborating with researchers to strengthen defenses.

The role involves regular exposure to potentially disturbing content and requires strong documentation, written communication, and ethical judgment. Seattle-based, contract, 40 hours/week.