Текст заказа
About The Role
Rex.zone is hiring Germany-based, bilingual English/German AI Generalist Trainers to support RLHF and large language model evaluation. You will assess, rank, QA, and validate model-generated outputs to strengthen training data quality and improve model performance across multilingual use cases.
Responsibilities
* Rate and rank model outputs in English and German using RLHF-style preference ranking
* Perform prompt evaluation and identify the best response based on helpfulness, correctness, and safety
* Conduct QA evaluation via consistency checks, edge-case reviews, and validation of annotations
* Write concise, evidence-based rationales explaining rankings and decisions
* Follow annotation guidelines, document exceptions, and propose updates for clarity and throughput
* Execute data labeling tasks, including content safety labeling when required
* Validate training data quality using checklists, rubrics, and targeted audits
* Escalate ambiguous cases and collaborate to resolve guideline gaps
Basic Qualifications
* Must be based in Germany and eligible to work remotely from Germany
* Fluent in English and German (C1/C2 or equivalent), strong reading and writing skills in both
* Strong analytical reasoning with ability to compare outputs and justify rankings
* High attention to detail and consistency when applying rubrics and annotation guidelines
* Comfortable with structured evaluation tasks, spreadsheets/tools, and written feedback loops
Preferred Qualifications
* Experience in data labeling, QA evaluation, prompt evaluation, or RLHF workflows
* Familiarity with LLM behavior and common failure modes
* Experience improving annotation guidelines and training data quality processes
* Self-driven, reliable, and effective time management in a remote, high-accuracy setting
Compensation
$30–$50 USD per hour (base pay; final rate based on skills and assessment performance).
How To Apply
Submit a short summary of your English/German language background and any experience in evaluation, QA, data labeling, or content review. Shortlisted candidates complete a structured assessment covering ranking, reasoning, and annotation guidelines compliance.