For employers

Hire this AI Trainer

Sign in or create an account to invite AI Trainers to your job.

Invite to Job
Ross Carmichael

Ross Carmichael

Senior AI Data Trainer - Technology & Internet

SOUTH_AFRICA flag
Cape Town, South Africa
$18.00/hrIntermediateInternal Proprietary Tooling

Key Skills

Software

Internal/Proprietary Tooling

Top Subject Matter

No subject matter listed

Top Data Types

ImageImage
TextText
VideoVideo

Top Label Types

Action Recognition
Emotion Recognition
Evaluation Rating
Fine Tuning
Object Detection
Prompt Response Writing SFT
Question Answering
RLHF
Text Generation
Text Summarization

Freelancer Overview

I am an experienced AI data trainer with a strong background in data annotation, quality assurance, and model evaluation. My work involves training AI systems in reading, writing, summarizing, and interpreting meaning, as well as annotating and reviewing video and image data to ensure high-quality outputs for model improvement. I am skilled in identifying inconsistencies, labeling data accurately, and applying detailed project guidelines to support the development of advanced AI, particularly in computer vision and natural language processing domains. My expertise with Google Workspace and Microsoft Office, combined with my attention to detail and clear communication, enables me to deliver reliable results in fast-paced, collaborative environments.

IntermediateEnglish

Labeling Experience

Prompt Response Annotations

Internal Proprietary ToolingImageEvaluation Rating
The campaign is a simple, and fast labeling pipeline that tracks the overall quality of responses on 10 critical dimensions. explicit instruction following implicit instruction following truthfulness harmlessness conciseness relevance content completeness insightfulness writing style and tone formatting overall quality

The campaign is a simple, and fast labeling pipeline that tracks the overall quality of responses on 10 critical dimensions. explicit instruction following implicit instruction following truthfulness harmlessness conciseness relevance content completeness insightfulness writing style and tone formatting overall quality

2025 - 2025

Faceswap Evals

Internal Proprietary ToolingImageEvaluation Rating
The overall goal is to replace the target image’s face and skin tone with the user’s face and skin tone while keeping the rest of the scene intact. The ideal FaceSwap image is one where the user’s face is naturally immersed into the target image’s face subject, matching the original image’s scene composition and lighting, the target face’s expression, eye gaze and positioning, and the user’s ID preservation. 9 Axis for Evaluation: ID Preservation Output Skin Tone Change Eye Gaze & Expression Match Style Transfer Physical Refitting Scene Preservation Hair Not Transferred Overall Assessment User Skin Tone Preservation

The overall goal is to replace the target image’s face and skin tone with the user’s face and skin tone while keeping the rest of the scene intact. The ideal FaceSwap image is one where the user’s face is naturally immersed into the target image’s face subject, matching the original image’s scene composition and lighting, the target face’s expression, eye gaze and positioning, and the user’s ID preservation. 9 Axis for Evaluation: ID Preservation Output Skin Tone Change Eye Gaze & Expression Match Style Transfer Physical Refitting Scene Preservation Hair Not Transferred Overall Assessment User Skin Tone Preservation

2025 - 2025

Image to Video Annotations

Internal Proprietary ToolingVideoText GenerationEmotion Recognition
Evaluated AI-generated videos created from source images and prompts, checking prompt accuracy, content preservation, visual quality, identity consistency, and the absence of artifacts or distortions.

Evaluated AI-generated videos created from source images and prompts, checking prompt accuracy, content preservation, visual quality, identity consistency, and the absence of artifacts or distortions.

2025 - 2025

Media Generation Prompt Reverse-Engineering

Internal Proprietary ToolingVideoQuestion AnsweringEvaluation Rating
Created mock user–AI conversations that simulate media generation (images, video, audio, etc.). Instead of generating new content, we reverse-engineered existing online media by writing the prompt that could have produced it. Each task included a real media example and a minimum four-turn dialogue where the user requests the media and the AI “generates” it.

Created mock user–AI conversations that simulate media generation (images, video, audio, etc.). Instead of generating new content, we reverse-engineered existing online media by writing the prompt that could have produced it. Each task included a real media example and a minimum four-turn dialogue where the user requests the media and the AI “generates” it.

2025 - 2025

Video Annotations

Internal Proprietary ToolingVideoText GenerationEmotion Recognition
Evaluated AI-generated video edits against an original source video and prompt, checking prompt accuracy, identity preservation, and visual quality (no artifacts, disfigurements, or lighting changes).

Evaluated AI-generated video edits against an original source video and prompt, checking prompt accuracy, identity preservation, and visual quality (no artifacts, disfigurements, or lighting changes).

2025 - 2025

Education

H

Hirt & Carter

Diploma, DTP / Mac Artworker / Graphic Design

Diploma
2000 - 2000

Work History

M

MTL Trading

Account Manager

Cape Town
2012 - Present
I

Invisible Technologies

Senior AI Data Trainer

San Francisco
2023 - 2025