Remote Data Labeling Specialist (Canada)

Remote Data Labeling Jobs in Canada (Full Time)

Rex.zone supports AI/ML training pipelines through data labeling, RLHF evaluation, prompt evaluation, and QA checks. You will apply annotation guidelines compliance to improve training data quality for large language models and computer vision systems.

About The Role You will label and evaluate text, image, and multimodal data used to train and validate machine learning models. Typical work includes LLM response grading, RLHF preference labeling, prompt evaluation, entity tagging for NLP, bounding boxes and segmentation for computer vision annotation, and content safety labeling. You will follow annotation guidelines, document edge cases, and complete QA evaluation checks to ensure dataset consistency and high training data quality.

Key Responsibilities Produce accurate labels for NLP and computer vision tasks

Perform RLHF ranking and pairwise preference judgments for LLM training

Execute prompt evaluation and rubric-based scoring for model outputs

Apply named entity recognition and taxonomy tagging

Complete content safety labeling with clear rationales

Run QA evaluation workflows including audits, cross-checks, and error analysis

Track annotation guidelines compliance and propose guideline improvements

Escalate ambiguous cases and contribute to calibration sessions that improve inter-annotator agreement

Required Qualifications Professional experience in data labeling, data annotation, QA evaluation, or trust and safety

Strong attention to detail and ability to follow annotation guidelines

Comfortable working with web-based annotation tools and spreadsheets

Ability to explain decisions clearly using rubrics, rationales, and examples

Familiarity with NLP concepts such as named entity recognition and text classification

Availability for full-time remote work with reliable connectivity

Preferred Qualifications Experience with RLHF workflows, LLM evaluation, and prompt evaluation

Exposure to computer vision annotation (bounding boxes, polygons, segmentation masks, keypoints)

Understanding of training data quality metrics (accuracy, consistency, coverage, bias)

Prior work with content safety labeling and policy interpretation

Experience collaborating with AI labs, tech startups, BPOs, or annotation vendors

Tools and Workflows You may use labeling platforms, internal QA dashboards, and guideline repositories. Workflows can include gold-standard calibration, blind reviews, inter-annotator agreement checks, and structured error analysis aimed at model performance improvement for production AI systems.

#J-18808-Ljbffr