AI Safety Expert, English – Bengali
Employment Information
Report this job
Job expired or something wrong with this job?
Job Description
AI red-teaming experts probing conversational models for Mercor, which helps AI labs train frontier systems. Testing vulnerabilities and producing reproducible safety data in English and Bengali.
Responsibilities:
- Red-team conversational AI models and agents using jailbreaks, prompt injections, misuse cases, bias exploitation, and multi-turn manipulation
- Generate human data by annotating failures, classifying vulnerabilities, and flagging systemic risks
- Apply taxonomies, benchmarks, and playbooks to maintain consistent testing
- Produce reproducible reports, datasets, and attack cases for customers
- Review AI outputs involving sensitive topics such as bias, misinformation, and harmful behaviors
- Expand evaluation coverage and uncover vulnerabilities missed by automated tests
Requirements:
- Native fluency in English and Bengali
- Strong judgment about language and content, including assessing whether AI responses are accurate, complete, and appropriate
- Ability to identify subtle errors, inconsistencies, and gaps
- Ability to consistently follow guidelines and quality standards
- Ability to explain reasoning clearly to technical and non-technical audiences
- Adaptability across projects, task types, and customers
- Independent contractor status
- Must not require H1-B or STEM OPT support
- Nice-to-have specialties: adversarial ML, cybersecurity, socio-technical risk, or creative probing
Benefits:
- Fully remote work
- Flexible own schedule
- Weekly payments via Stripe or Wise
- Projects may be extended, shortened, or concluded early depending on needs and performance
- Higher-sensitivity project participation is optional
- Clear guidelines and wellness resources for sensitive-content work
- Reasonable accommodations upon request
- Referral payments of up to $90 per successful referral
Similar Jobs
AI Safety Expert – English, Telugu
AI safety red teamer probing conversational models for Mercor’s human-data AI training projects. Generating reproducible vulnerability data to strengthen frontier AI systems.
AI Safety Red Teamer
AI Safety Red Teamer stress-testing frontier AI models for Mercor, which partners with AI labs to train and enhance AI systems. Identifying jailbreaks, safety failures, and vulnerabilities.
AI Safety Expert – English, Assamese
AI red-teaming expert probing conversational models for Mercor, which partners with AI labs to train frontier systems. Generating adversarial data, vulnerability classifications, and reproducible attack artifacts.
AI Safety Expert, English, Odia
AI safety red teamers probing conversational AI models for vulnerabilities. Generating human data and reproducible attack artifacts for Mercor’s AI lab and enterprise customers.
Accounting, Data, and Artificial Intelligence Internship
Estagiário de Contabilidade preparando dados financeiros para IA e automação. Instituto Infnet, faculdade brasileira focada em tecnologia, dados, comunicação e design.
AI Safety Red Teamer
AI Safety Red Teamer identifying vulnerabilities in frontier AI systems for Mercor. Designing adversarial prompts and evaluating high-risk model behavior.
AI Safety Red Teamer
AI Safety Red Teamer stress-testing frontier AI models for Mercor. Identifying vulnerabilities and improving model alignment through adversarial evaluation.
AI Safety Experts – English, Odia
AI safety expert red-teaming conversational AI models for Mercor’s frontier-model training projects. Generating vulnerability annotations, attack cases, datasets, and reproducible safety reports.
AI Safety Red Teamer
AI Safety Red Teamer stress-testing frontier AI models for Mercor. Identifying vulnerabilities and improving model alignment, robustness, and safety through adversarial evaluation.
AI Safety Red Teamer
AI Safety Red Teamer identifying vulnerabilities in frontier AI models for Mercor’s AI training projects. Designing adversarial prompts and evaluating high-risk AI behavior.
AI Safety Expert, English – Kannada
AI red team expert probing Mercor’s frontier models for vulnerabilities using adversarial English and Kannada inputs. Generating reproducible safety data, reports, and attack cases for AI customers.
AI Safety Expert – English, Punjabi
AI red-teaming contractor probing conversational models for Mercor’s human-data AI safety projects. Generating vulnerability annotations, attack cases, datasets, and reproducible safety reports.
AI Safety Expert – English, Tamil
AI red-team expert probing conversational models for Mercor’s human-powered frontier AI training. Testing vulnerabilities, annotating failures, and producing reproducible safety datasets.
AI Safety Expert – English, Tamil
AI red-team contractor probing conversational models for Mercor, which helps AI labs and enterprises train frontier systems. Testing jailbreaks, vulnerabilities, and harmful behaviors through structured human data.
AI Safety Red Teamer
AI Safety Red Teamer stress-testing frontier AI models for Mercor. Identifying vulnerabilities and improving model alignment, robustness, and safety through adversarial evaluation.
AI Safety Expert, English – Tamil
AI red-teaming experts probing conversational models for Mercor, which partners with AI labs to train frontier systems. Testing vulnerabilities and generating actionable human safety data.
AI Safety Expert – English, Gujarati
AI red team expert probing conversational models for Mercor, which partners with AI labs to train frontier systems. Generating adversarial datasets, vulnerability reports, and attack cases.
AI Safety Red Teamer
AI Safety Red Teamer stress-testing Mercor's frontier AI models. Identifying vulnerabilities across high-risk domains and improving model alignment, robustness, and safety.
AI Safety Expert, English, Telugu
AI red-teaming expert probing conversational models for Mercor, which partners with AI labs to train frontier systems. Generating vulnerability data, attack cases, and reproducible safety reports.
AI Safety Expert, English – Kannada
AI red-team expert probing frontier models for Mercor’s human-data AI training projects. Testing vulnerabilities and generating reproducible datasets, reports, and attack cases.