Job Description
Note: The job is a remote job and is open to candidates in USA. Rex.zone is seeking individuals for entry level AI roles to support human-in-the-loop AI training workflows for large language models and multimodal systems. The role involves improving training data quality through data labeling, preference ranking, and various evaluation tasks using structured tools and guidelines.
Responsibilities
- Create and review labeled datasets for NLP, computer vision, and multimodal use cases
- Perform RLHF preference ranking and rubric-based scoring
- Run prompt and response evaluations for helpfulness, factuality, and safety
- Execute content safety labeling (policy, harassment, self-harm, sensitive categories)
- Follow annotation guidelines and document rationales for edge cases
- Conduct QA evaluation via sampling plans, audits, and error taxonomy tracking
- Participate in calibration sessions to reduce rater variance and improve agreement
- Track throughput and quality metrics that impact LLM training pipelines
Skills
- Strong written English and structured reasoning
- Ability to follow detailed annotation guidelines with consistent judgment
- Comfort using spreadsheets, web tools, or labeling interfaces
- Familiarity with model evaluation and prompt/response patterns
- Attention to detail and evidence-based QA evaluation approach
- Comfort handling sensitive content as part of content safety labeling
Company Overview
Apply To This Job