Portuguese Data Trainer - Remote

About the role

This remote, hourly‑paid contractor position focuses on improving the safety, reliability, and reasoning quality of advanced AI systems. You will evaluate AI‑generated content in Brazilian Portuguese and English, applying linguistic, cultural, and policy judgment to ensure outputs are accurate, logical, and aligned with safety standards.

Key responsibilities

• Review AI‑generated responses for accuracy, clarity, reasoning quality, and compliance with safety policies.
• Identify logical, methodological, or conceptual errors in model outputs.
• Compare and rate multiple AI responses based on correctness, safety alignment, and policy adherence.
• Perform fact‑checking and validation where required.
• Conduct red‑teaming and adversarial testing to uncover vulnerabilities, edge cases, and unsafe behaviors.
• Document findings with clear, structured reasoning and reproducible decision paths.
• Apply and localize safety policies across Portuguese and English, recognizing slang, coded language, and cultural nuance.
• Escalate ambiguous cases using defined policy frameworks.

Required profile

• Bachelor’s degree or higher in Linguistics, Communications, Psychology, Law, Policy, Security Studies, or related field (or equivalent experience).
• Native or near‑native Brazilian Portuguese proficiency (reading and writing).
• Minimum C1 English proficiency (reading and writing).
• 5+ years of experience in Trust & Safety, content moderation, policy operations, compliance, investigations, or related functions.
• Proven LLM red‑teaming / adversarial testing experience.
• Strong analytical writing and structured reasoning skills.
• High attention to detail and consistency in applying safety guidelines.

Required skills

• LLM red‑teaming
• Adversarial testing
• ChatGPT
• Gemini
• Perplexity