Remote Data Labeling Jobs Copenhagen

Rex.zone

Fjernarbejde
Overview

Rex.zone connects candidates to remote AI data operations teams building training and evaluation datasets for modern AI/ML systems. This role focuses on high-quality annotation and evaluation work that improves large language model evaluation outcomes and model performance.

What You Will Do

  • Produce accurate labels at scale across NLP and computer vision tasks
  • Maintain annotation guidelines compliance and document edge-case decisions
  • Complete RLHF comparisons and preference ranking using rubrics
  • Perform QA evaluation using checklists, sampling, and gold-set reviews
  • Identify ambiguity, report issues, and propose guideline updates
  • Support content safety labeling and confidentiality requirements in a remote setting

Common Task Types

  • Prompt evaluation and instruction-following checks for chat models
  • Named entity recognition, sentiment/intent classification, and policy labeling
  • Computer vision annotation (bounding boxes, polygons, keypoints) and OCR verification
  • Dataset refresh cycles, adjudication support, and error tracking/rework

Required Qualifications

  • Mid-senior experience in data labeling, data annotation, QA, or evaluation workflows
  • Strong reading comprehension and consistent written reasoning for rubric-based scoring
  • Familiarity with NLP tasks and/or computer vision annotation
  • Ability to apply content safety policies consistently

Preferred Qualifications

  • Experience with RLHF pipelines, prompt evaluation, and LLM evaluation frameworks
  • Exposure to inter-annotator agreement, calibration, and adjudication workflows
  • Comfort working across multiple domains and task types

Workplace

This is a Remote role. “Copenhagen” reflects search intent and location-based discovery; work is performed remotely with distributed teams.

Compensation

Hourly base pay range: $30–$50/hr (varies by project and scope).

Hvordan man ansøger

For at ansøge om dette job skal du autorisere på vores websted. Hvis du ikke har en konto endnu, bedes du tilmelde.