Overview
Vezloo is seeking a Prompt Engineer specializing in adversarial testing of Generative AI systems. This role focuses on designing and executing structured prompt-based attack scenarios to evaluate model safety, robustness, and compliance across multiple risk domains. You will contribute to strengthening AI systems by identifying vulnerabilities and documenting model behavior under challenging inputs.
Key Responsibilities
- Design and develop adversarial prompts across multiple risk categories, including but not limited to:
- Harmful or unsafe content generation
- Bias and fairness issues
- Misinformation and factuality manipulation
- Privacy leakage and sensitive data exposure
- Policy bypass and safety guardrail evasion
- Execute structured attack scenarios against GenAI systems, such as:
- Prompt injection techniques
- Jailbreaking attempts
- Multi-turn conversational manipulation
- Role-play and context distortion strategies
- Analyze model outputs to identify failures, inconsistencies, or policy violations
- Document findings clearly with actionable insights for model improvement
- Collaborate with evaluation teams to refine testing frameworks and coverage strategies
Requirements
- English proficiency level C1 or C2 (written and spoken)
- Proven experience in Generative AI model evaluation or assessment
- Strong hands-on experience in prompt engineering and structured prompt design
- Familiarity with adversarial testing concepts in AI systems
- Ability to think analytically and simulate edge-case user behaviors
- Access to a Mac device running the latest macOS
Benefits
- Fully remote work environment
- Opportunity to work on cutting-edge AI technologies
- Collaborative and inclusive team culture
- Professional growth and development opportunities
