Cultural à Brésil

Evaluate outputs generated by LLMs across multiple modalities (text, image captions, video descriptions, and multimodal prompts).

Assess quality against project-specific criteria such as correctness, coherence, completeness, style, cultural appropriateness, and safety.

Identify subtle errors, hallucinations, or biases in AI responses.

Apply domain expertise and logical reasoning to resolve ambiguous or unclear outputs.

Provide detailed written feedback, tagging, and scoring of outputs to ensure consistency across the evaluation team.

Escalate unclear cases and contribute to refining evaluation guidelines.

Collaborate with Project Managers and Quality Leads to meet accuracy, reliability, and turnaround benchmarks.

Strong critical reading, observational, and evaluative skills across different modalities.

Ability to articulate nuanced judgments with precision and clarity.

Excellent English comprehension (CEFR B2 or above); additional languages a plus.

Familiarity with LLMs, generative AI, and multimodal systems.

Strong attention to detail and ability to apply guidelines consistently.

Awareness of cultural and linguistic nuances, including potential bias and harm in AI outputs.

Comfort with evolving workflows, rapid feedback cycles, and complex quality frameworks.

Bachelor's degree / diploma or equivalent educational qualification.

1+ years of experience in data annotation, LLM evaluation, content moderation, or related AI / ML domains.

Demonstrated experience working with data annotation tools and software platforms.

Strong understanding of language and multimodal communication (instruction following in image generation, fact-checking, narrative coherence in video, etc.).

Ability to adapt quickly to changing project directions and fast-paced work environments.

Previous experience creating or annotating complex data specifically for Large Language Model (LLM) training.

Prior exposure to generative AI, prompt engineering, or LLM fine-tuning workflows is a plus.

While moderation of high-harm / high-risk material is not part of this role, candidates should be aware that occasional exposure to NSFW or otherwise sensitive content may occur due to imperfections in client-provided datasets. Applicants should indicate that they are comfortable working in environments where such incidental exposure is a possibility.

Opportunities to shape the evaluation standards for next-generation multimodal AI systems.

Innovative and supportive global working environment.

Competitive compensation and flexible remote working arrangements.

Join iMerit to help shape the benchmarks and standards that ensure AI systems are accurate, safe, and culturally aware across text, vision, and multimodal applications. If rigorous analysis and innovation excite you, we encourage you to apply!