Adversarial Tester for Large Language Models
Budget: $15 – $25 USD
We're looking for an adversarial tester based in India or the US who specializes in breaking and “failing” large language model (LLM) systems through creative prompt engineering, red‑teaming, and stress testing. The goal of this role is to systematically expose safety, reliability, and robustness gaps so they can be measured, fixed, and prevented in future releases.
The role involves designing and executing adversarial prompt campaigns to deliberately expose failures in LLMs, discovering and documenting new failure modes, and building test suites to track model robustness over time. It also includes performing red-team exercises, analyzing model responses at scale, collaborating with teams to implement mitigations, and producing clear reports with findings and recommendations. Additionally, the tester will contribute to internal knowledge bases and best practices for adversarial testing and prompt engineering.
Candidates should have hands-on experience with LLMs and adversarial prompt techniques (e.g., jailbreaks, prompt injection), a strong understanding of LLM architectures, and the ability to think creatively and systematically to identify edge cases. They should also be skilled in designing experiments, running structured tests, analyzing outputs (e.g., with Python), and clearly documenting findings for technical and non-technical audiences. Preferred qualifications include experience in security red‑teaming, AI safety, integrating adversarial tests into CI/CD pipelines, and knowledge of responsible AI and privacy principles.
Pay Rates
India: $14/hr
USA: Up to $40/hr
The role involves designing and executing adversarial prompt campaigns to deliberately expose failures in LLMs, discovering and documenting new failure modes, and building test suites to track model robustness over time. It also includes performing red-team exercises, analyzing model responses at scale, collaborating with teams to implement mitigations, and producing clear reports with findings and recommendations. Additionally, the tester will contribute to internal knowledge bases and best practices for adversarial testing and prompt engineering.
Candidates should have hands-on experience with LLMs and adversarial prompt techniques (e.g., jailbreaks, prompt injection), a strong understanding of LLM architectures, and the ability to think creatively and systematically to identify edge cases. They should also be skilled in designing experiments, running structured tests, analyzing outputs (e.g., with Python), and clearly documenting findings for technical and non-technical audiences. Preferred qualifications include experience in security red‑teaming, AI safety, integrating adversarial tests into CI/CD pipelines, and knowledge of responsible AI and privacy principles.
Pay Rates
India: $14/hr
USA: Up to $40/hr