Karen Dobbelstein

Key Skills

Software

Other

Scale AI

Internal/Proprietary Tooling

Top Subject Matter

No subject matter listed

Top Data Types

Text

Image

Top Label Types

Evaluation Rating

Bounding Box

Red Teaming

Classification

Object Detection

Freelancer Overview

I am a language specialist with over 13 years of experience evaluating and refining written content, with a strong recent focus on generative AI training data, dialogue evaluation, and rubric-based quality assessment. My work includes scoring AI-generated responses for clarity, accuracy, tone, and instruction adherence, as well as identifying edge cases and providing structured feedback to improve model alignment and linguistic precision. I am skilled in data annotation across text, image, audio, and video, and have hands-on experience with AI training platforms, detailed rubric interpretation, and error classification. My background in writing, editing, and instructional design enables me to deliver high-quality, consistent feedback in production environments, and I am continuously upskilling in AI evaluation methodologies and quality frameworks to stay at the forefront of the field.

Labeling Experience

Freelance AI Safety Evaluator (Red Teaming)

Internal Proprietary ToolingTextRed Teaming

Conducted red teaming and AI safety evaluation tasks focused on identifying failure modes in generative model behavior. Work involved crafting adversarial text prompts, testing boundary conditions, and documenting model responses using structured feedback fields rather than traditional annotation labels. Outputs included categorization of failure types, severity assessments, and concise written rationales, all completed within client-provided proprietary evaluation platforms and aligned with project quality and safety guidelines.

Conducted red teaming and AI safety evaluation tasks focused on identifying failure modes in generative model behavior. Work involved crafting adversarial text prompts, testing boundary conditions, and documenting model responses using structured feedback fields rather than traditional annotation labels. Outputs included categorization of failure types, severity assessments, and concise written rationales, all completed within client-provided proprietary evaluation platforms and aligned with project quality and safety guidelines.

2025

Freelance Image Classification Annotator

OtherImageClassificationObject Detection

Completed basic image classification tasks involving identification and selection of specified objects within images. Work required accurate recognition of visual elements and adherence to simple labeling guidelines to support training data quality.

Completed basic image classification tasks involving identification and selection of specified objects within images. Work required accurate recognition of visual elements and adherence to simple labeling guidelines to support training data quality.

2025 - 2025

Freelance Annotator

Scale AIImageBounding Box

Performed image-based data annotation involving bounding box labeling for visual elements in charts and tables, including graphs, pie charts, axes, legends, labels, and tabular components. Tasks required identifying, classifying, and accurately enclosing all relevant elements according to project-specific guidelines to ensure complete coverage and consistency. Adhered to quality standards for precision, completeness, and label accuracy to support reliable downstream model training.

Performed image-based data annotation involving bounding box labeling for visual elements in charts and tables, including graphs, pie charts, axes, legends, labels, and tabular components. Tasks required identifying, classifying, and accurately enclosing all relevant elements according to project-specific guidelines to ensure complete coverage and consistency. Adhered to quality standards for precision, completeness, and label accuracy to support reliable downstream model training.

2025 - 2025

Freelance Annotator

OtherTextEvaluation Rating

Conducted generative AI output evaluations using multi-dimensional, complex rubrics on the Outlier (Scale AI) platform. The project involved assessing model responses across 3-4+ prioritized criteria per task, each scored on a 1-5 scale with required written justifications. Evaluations focused on accuracy, instruction-following, clarity, tone, and overall response quality. Adhered to strict quality and consistency standards, including rubric alignment, evidence-based scoring, and detailed reasoning to support ratings. Work required careful calibration to ensure inter-rater reliability and identification of rubric edge cases or ambiguities.

Conducted generative AI output evaluations using multi-dimensional, complex rubrics on the Outlier (Scale AI) platform. The project involved assessing model responses across 3-4+ prioritized criteria per task, each scored on a 1-5 scale with required written justifications. Evaluations focused on accuracy, instruction-following, clarity, tone, and overall response quality. Adhered to strict quality and consistency standards, including rubric alignment, evidence-based scoring, and detailed reasoning to support ratings. Work required careful calibration to ensure inter-rater reliability and identification of rubric edge cases or ambiguities.

2025 - 2025

Education

Work History