AI Safety Expert for English and Bengali Red Teaming

Lead adversarial testing of conversational AI by probing models with targeted, text-based attacks, surfacing vulnerabilities, and producing reproducible red team data that helps customers harden their systems. This role focuses on reviewing model outputs on sensitive topics such as bias, misinformation, and harmful behaviors, with explicit safeguards for higher-sensitivity work.

• Perform red teaming of conversational AI and agents, including jailbreaks, prompt injection, misuse scenarios, bias exploitation, and multi-turn manipulation.
• Review AI outputs that touch on sensitive topics, annotate failures, classify vulnerabilities, and flag systemic risks.
• Follow prescribed taxonomies, benchmarks, and playbooks to keep testing consistent and structured.
• Produce reproducible documentation and deliverables, such as reports, labeled datasets, and concrete attack cases customers can act on.
• Opt in to or out of higher-sensitivity projects, with topics communicated ahead of exposure and access to clear guidelines and wellness resources when participating.

• Native fluency in English and Bengali, both written and spoken, is required.
• Prior red teaming experience, such as AI adversarial work, cybersecurity, or socio-technical probing.
• Comfort with adversarial, curiosity-driven approaches, pushing systems toward failure modes.
• Ability to work with structured frameworks, not rely on random or ad hoc techniques.
• Strong communication skills, able to explain risks to both technical and non-technical stakeholders.
• Adaptable to move between different projects and customer contexts.

• Adversarial ML experience, for example jailbreak datasets, prompt injection, RLHF or DPO attack methods, model extraction.
• Cybersecurity skills such as penetration testing, exploit development, or reverse engineering.
• Socio-technical risk expertise, including harassment, disinformation probing, abuse analysis, or conversational AI testing.
• Creative probing abilities drawing on psychology, acting, or creative writing for unconventional adversarial thinking.

• Identifies vulnerabilities that automated tests miss.
• Delivers reproducible artifacts that customers use to strengthen AI systems.
• Expands evaluation coverage so more scenarios are tested and fewer surprises occur in production.
• Increases customer confidence in the safety of their deployed AI through thorough adversarial testing.

• Location: Remote.
• Employment type: hourly engagement.
• All tasks are text-based; no audio or video review is required.

• Pay rate: 20 - 22 hourly.

• Native-level fluency in English and Bengali is mandatory.
• Participation in higher-sensitivity content is optional and will be preceded by clear communication of topics and supported by guidelines and wellness resources.