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