AI Jailbreak & Prompt-Injection Security Expert, Remote- Contract Remote (International)

Xperteez Technology Pvt Ltd; India; A distancia; INR 7.880.000 - 15.759.000; A tiempo parcial

Descripción de la vacante

Dover is seeking an AI Jailbreak & Prompt-Injection Security Expert for a remote contract role. You will design and test advanced safety assessments for large language models, focusing on ethical jailbreaks, red teaming, prompt injection, and tool-use abuse.

Ideal candidates have 2+ years in adversarial ML or AI safety evaluation, with an advanced degree or equivalent experience, and strong ability to document findings for technical and non-technical stakeholders.

Formación

• 2+ years of expertise in adversarial ML, red teaming or AI safety evaluation.
• Proven experience researching ethical jailbreaks, prompt injection, tool-use abuse, or adversarial AI attacks.
• Advanced degree (PhD or MS) in CS, cybersecurity, ML, or related field, or equivalent experience.
• Strong written and verbal communication with clear documentation.

Responsabilidades

• Design methods to evaluate AI safety including ethical jailbreaks and prompt-injection scenarios.
• Create cross-domain elicitation strategies to uncover multi-turn bypass patterns.
• Develop and maintain regression test suites for jailbreak susceptibility and prompt-injection vulnerabilities.
• Build evaluation frameworks to stress-test models against real-world adversarial threats.
• Collaborate with stakeholders to translate findings into actionable model improvements.
• Document methodologies and findings in clear reports for technical and non-technical audiences.

Conocimientos

Ethical Jailbreaks

LLM Red Teaming

Prompt Injection

Tool-Use Abuse

Educación

Advanced degree (PhD, MS)

Descripción del empleo

Role Title: AI Jailbreak & Prompt-Injection Security Expert

Role Type: Contract

Location: Remote

Required Skills:

• Ethical Jailbreaks
• LLM Red Teaming
• Prompt Injection
• Tool-Use Abuse

In this role, you'll apply your expertise to help train next-generation AI systems. Your work will shape how models learn, reason, and perform through high-quality, real-world input. No prior experience in AI is required - your domain knowledge is what matters.

Scope of Work:

• Design and implement advanced methodologies for evaluating AI system safety, focusing on ethical jailbreaks, LLM red teaming, prompt injection, and tool-use abuse scenarios.
• Create comprehensive cross-domain elicitation strategies to uncover multi-turn and complex adversarial bypass patterns in AI models.
• Develop, maintain, and update regression test suites that systematically test for jailbreak susceptibility and prompt-injection vulnerabilities.
• Construct robust evaluation frameworks that stress-test AI models against real-world adversarial threats, aiming to enhance overall system robustness.
• Collaborate with technical stakeholders to translate security findings into actionable improvements for model safety and risk mitigation.
• Document methodologies, findings, and best practices in clear, well-structured written reports and presentations for both technical and non-technical audiences.

Preferred Qualifications:

• 2+ years of expertise in adversarial machine learning, LLM red teaming, AI safety evaluation, or a closely related security domain
• Proven experience researching, testing, or uncovering vulnerabilities related to ethical jailbreaks, prompt injection, tool-use abuse, or adversarial AI attacks.
• Advanced degree (PhD, MS) in computer science, cybersecurity, machine learning, or a relevant discipline, or equivalent operational/professional background.
• High credibility and recognition within the AI security or adversarial ML community - such as published research, open-source tools, or conference presentations.
• Exceptional written and verbal communication skills, with a strong focus on clear documentation and collaborative problem-solving.
• Prior participation in multi-disciplinary projects or cross-functional AI safety initiatives is a plus.
• Familiarity with current LLM architectures, prompt engineering techniques, and security assessment tools is highly desirable.