AI Safety Experts — English & Odia

📋 Description

• Red-team conversational AI models and agents by testing jailbreaks, prompt injections, misuse
• Identify vulnerabilities and safety gaps that automated testing may miss.
• Annotate model failures and document adversarial examples to create high-quality human data.
• Apply taxonomies, benchmarks, and playbooks for consistent evaluations.
• Create structured attack cases, datasets, and reports to improve AI safety and model performance.
• Communicate technical and behavioral risks to technical and non-technical stakeholders.

🎯 Requirements

• Native-level fluency in English and Odia is required.
• Experience with AI red teaming, adversarial AI, cybersecurity, socio-technical probing, AI
• Strong curiosity and adversarial mindset to identify manipulation risks.
• Systematic testing using frameworks, benchmarks, taxonomies, and structured methodologies.
• Strong written and verbal communication to explain vulnerabilities and findings clearly.
• Excellent analytical, problem-solving, and documentation skills.

🎁 Benefits

• Fully remote work with flexible project timelines.
• Independent contractor engagement with weekly payments via Stripe or Wise based on services
• Competitive compensation based on expertise and assignment.
• Opportunity to work on cutting-edge AI safety and human-data projects.
• Direct contribution to more robust, safe, trustworthy AI systems.
• Exposure to evolving AI evaluation and red-teaming methodologies.