Kathy Cookson

Key Skills

Software

Scale AI

Internal/Proprietary Tooling

Top Subject Matter

No subject matter listed

Top Data Types

Audio

Image

Text

Video

Top Label Types

Evaluation Rating

Fine Tuning

Object Detection

Prompt Response Writing SFT

RLHF

Tracking

Freelancer Overview

I am an experienced AI Data Annotation Specialist with a strong background in delivering high-quality labeled datasets for large language models and multimodal AI systems. My expertise includes annotation, prompt writing, response review, and preference ranking of outputs across text, images, audio, and video. I consistently exceed client quality metrics, maintaining a 94%+ accuracy rate and an exceptionally low error rate of 1–4%, far surpassing industry standards. I am skilled at using specialized labeling platforms and applying systematic evaluation techniques to ensure data relevance, coherence, and safety. My work supports large-scale training data workflows, preference-based reinforcement learning, and human feedback loops, enabling the development and fine-tuning of robust AI models. Continuing education courses include: Rubrics, Prompt Engineering including Model Stumps, and Agent Coding.

Labeling Experience

Multimango

Internal Proprietary ToolingImageEvaluation Rating

Multi-platform, large project for human-in-the-loop AI development, focusing on evaluating, reviewing, and generating data through text, image, video, and audio. Tasks vary and include H2H preference ranking, prompt engineering for images and video reconstructions, bounding boxes, and object tracking and identification.

Multi-platform, large project for human-in-the-loop AI development, focusing on evaluating, reviewing, and generating data through text, image, video, and audio. Tasks vary and include H2H preference ranking, prompt engineering for images and video reconstructions, bounding boxes, and object tracking and identification.

2025

Steerability project

Scale AITextRLHFPrompt Response Writing SFT

Delivered 94%+ client quality metrics and an average error rate of only 1%−4% (versus 10% industry standard) for over one year as a team member on one of the highest-performing long-term steerability projects for the client. Tasks utilized RLHF techniques to steer bots within safety guidelines with a goal of appropriate, creative interactions with human participants. Prompt and Response writing was also used when needed to realign a bot to character or maintain safety standards.

Delivered 94%+ client quality metrics and an average error rate of only 1%−4% (versus 10% industry standard) for over one year as a team member on one of the highest-performing long-term steerability projects for the client. Tasks utilized RLHF techniques to steer bots within safety guidelines with a goal of appropriate, creative interactions with human participants. Prompt and Response writing was also used when needed to realign a bot to character or maintain safety standards.

2024 - 2025

Education

Work History