Arabic-Speaking AI Session Annotator (Voice & Vision QA Reviewer)

About the role

We're looking for detail-oriented, Arabic-fluent reviewers to help evaluate real interactions between users and a multimodal AI assistant. In these sessions, a person speaks aloud to the assistant while their camera streams whatever is happening in front of them — the assistant has to understand both what it hears and what it sees. Your job is to watch these recorded sessions and judge whether the assistant actually got it right.

This isn't repetitive labeling work. Every session requires close attention and independent judgment: you'll compare what the assistant claimed or said against what was truly visible in the video, then score the interaction against a structured rubric and write clear, specific feedback. Your written comments become the primary signal the AI training team uses to identify and prioritize failures — so precision and honesty in your annotations matter more than speed.

What you'll do

• Watch recorded voice-and-camera sessions between a user and an AI assistant from start to finish.
• Carefully compare the assistant's spoken responses against the actual visual content captured on camera during the session.
• Score each session against a detailed evaluation rubric covering accuracy, grounding, helpfulness, and safety.
• Write clear, specific, and constructive comments explaining any errors, hallucinations, or missed visual cues.
• Flag high-severity failures (factual errors, unsafe suggestions, or clear mismatches between what was said and what was shown) for priority review.
• Maintain consistency and calibration across sessions by applying rubric guidelines uniformly.
• Communicate ambiguous or edge-case scenarios to the team for rubric clarification when needed.

Requirements

• Native or near-native fluency in Arabic (reading, writing, and listening comprehension), with strong English proficiency for rubric documentation and internal communication.
• Excellent attention to detail and the ability to spot subtle inconsistencies between audio and visual information.
• Strong written communication skills — you'll be producing structured, actionable feedback, not just numeric scores.
• Comfort working independently with a rubric-based evaluation framework.
• Reliable internet connection and a quiet environment suitable for reviewing audio/video content.
• Prior experience in content review, QA, linguistics, translation, or AI/data annotation is a plus but not required.

Compensation

This role pays $17–$23 per hour depending on experience and evaluation quality, fully remote, with flexible scheduling. Payment is based on completed and approved session reviews.