AI Safety Experts
AI Safety Expert
This role focuses on strengthening the safety and reliability of advanced AI systems through structured adversarial testing and evaluation. You will work on identifying vulnerabilities in conversational models by designing and executing red-team scenarios in both English and Odia. The position involves analyzing AI behavior under edge cases such as bias, misinformation, and manipulation risks, helping to improve system robustness. You will contribute to building high-quality datasets that directly support AI safety research and model alignment. The work is highly analytical, language-driven, and requires both creativity and structured reasoning. It is fully remote, flexible, and suited for individuals who enjoy exploring system weaknesses to make AI safer and more trustworthy at scale.
Accountabilities:
• Design and execute adversarial test cases to evaluate AI model behavior, including jailbreaks, prompt injections, and multi-turn manipulation scenarios.
• Identify, classify, and document vulnerabilities, failures, and risk patterns in AI-generated outputs.
• Generate structured datasets and annotated examples that support AI safety evaluation and training.
• Follow defined taxonomies, benchmarks, and safety frameworks to ensure consistent evaluation standards.
• Produce clear, reproducible reports describing observed risks and system weaknesses.
• Collaborate on improving evaluation coverage by expanding test scenarios across linguistic and cultural contexts (English and Odia).
Requirements:
• Strong fluency in both English and Odia (native or near-native proficiency required).
• Prior experience in AI red teaming, cybersecurity, adversarial testing, or socio-technical risk analysis.
• Ability to think creatively and critically to identify system weaknesses and edge-case behaviors.
• Strong structured thinking skills, with the ability to follow frameworks and document findings clearly.
• Excellent written communication skills for both technical and non-technical audiences.
• Familiarity with AI systems, conversational agents, or machine learning concepts is a plus.
• Ability to work independently in a remote, task-driven environment.
Benefits:
• Competitive hourly compensation of 20 to 22 USD per hour.
• Fully remote and flexible work arrangement with independent scheduling.
• Opportunity to contribute directly to cutting-edge AI safety and alignment research.
• Exposure to advanced red-teaming methodologies and frontier AI systems.
• Weekly payments via Stripe or Wise as an independent contractor.
• Ongoing project extensions based on performance and impact.
This role focuses on strengthening the safety and reliability of advanced AI systems through structured adversarial testing and evaluation. You will work on identifying vulnerabilities in conversational models by designing and executing red-team scenarios in both English and Odia. The position involves analyzing AI behavior under edge cases such as bias, misinformation, and manipulation risks, helping to improve system robustness. You will contribute to building high-quality datasets that directly support AI safety research and model alignment. The work is highly analytical, language-driven, and requires both creativity and structured reasoning. It is fully remote, flexible, and suited for individuals who enjoy exploring system weaknesses to make AI safer and more trustworthy at scale.
Accountabilities:
• Design and execute adversarial test cases to evaluate AI model behavior, including jailbreaks, prompt injections, and multi-turn manipulation scenarios.
• Identify, classify, and document vulnerabilities, failures, and risk patterns in AI-generated outputs.
• Generate structured datasets and annotated examples that support AI safety evaluation and training.
• Follow defined taxonomies, benchmarks, and safety frameworks to ensure consistent evaluation standards.
• Produce clear, reproducible reports describing observed risks and system weaknesses.
• Collaborate on improving evaluation coverage by expanding test scenarios across linguistic and cultural contexts (English and Odia).
Requirements:
• Strong fluency in both English and Odia (native or near-native proficiency required).
• Prior experience in AI red teaming, cybersecurity, adversarial testing, or socio-technical risk analysis.
• Ability to think creatively and critically to identify system weaknesses and edge-case behaviors.
• Strong structured thinking skills, with the ability to follow frameworks and document findings clearly.
• Excellent written communication skills for both technical and non-technical audiences.
• Familiarity with AI systems, conversational agents, or machine learning concepts is a plus.
• Ability to work independently in a remote, task-driven environment.
Benefits:
• Competitive hourly compensation of 20 to 22 USD per hour.
• Fully remote and flexible work arrangement with independent scheduling.
• Opportunity to contribute directly to cutting-edge AI safety and alignment research.
• Exposure to advanced red-teaming methodologies and frontier AI systems.
• Weekly payments via Stripe or Wise as an independent contractor.
• Ongoing project extensions based on performance and impact.