John Gichana

Key Skills

Software

Data Annotation Tech

Mercor

OneForma

Remotasks

Telus

Scale AI

Top Subject Matter

No subject matter listed

Top Data Types

Audio

Image

Video

Top Label Types

Action Recognition

Cuboid

Evaluation Rating

Object Detection

Question Answering

Freelancer Overview

I bring over 5 years of experience in creative design, visual communication, and multimedia production, combined with hands-on expertise in AI data labeling and model evaluation. As an AI Assistant Evaluator at Prolific,Telus, Outlier/Remotask and mercor I contributed to refining large language models by assessing response quality, factual accuracy, and creative reasoning. My background in photography, animation, and design allows me to analyze complex visual data with a trained artistic eye, ensuring precise annotations, consistent labeling standards, and a deep understanding of how AI systems interpret visual context.

Labeling Experience

AI assistant evaluator

Data Annotation TechVideoRelationship

To evaluate AI-generated videos for their adherence to physical commonsense using a structured three-part approach Part 1: Rule Evaluation: Judge the video against five specific physical rules provided. Part 2: Additional Violations: Identify and describe any other physical-rule violations you observe. Part 3: Holistic Judgments & Quality Checks: Provide overall ratings for physical commonsense and prompt alignment, and flag any fundamental issues with the item or your ability to evaluate it.

To evaluate AI-generated videos for their adherence to physical commonsense using a structured three-part approach Part 1: Rule Evaluation: Judge the video against five specific physical rules provided. Part 2: Additional Violations: Identify and describe any other physical-rule violations you observe. Part 3: Holistic Judgments & Quality Checks: Provide overall ratings for physical commonsense and prompt alignment, and flag any fundamental issues with the item or your ability to evaluate it.

2025 - 2025

AI Assistant Evaluator

Scale AITextText Generation

To evaluate the quality of an Expanded Prompt. This process starts with a short User Prompt (a user's request for an image), which a language model then expands into a longer, more detailed prompt. The goal is to enrich the original request with helpful, accurate, and consistent details to prepare it for an image generation model.

To evaluate the quality of an Expanded Prompt. This process starts with a short User Prompt (a user's request for an image), which a language model then expands into a longer, more detailed prompt. The goal is to enrich the original request with helpful, accurate, and consistent details to prepare it for an image generation model.

2025 - 2025

Evaluate an AI Assistant - Compare two conversations

Data Annotation TechTextRelationship

Rate conversations between a User and an AI assistant and judge how good the assistant responses are.

Rate conversations between a User and an AI assistant and judge how good the assistant responses are.

2025

AI Videos - Evaluation

MercorVideoRelationship

Each question contains two media samples and a question asking you to pick the best one according to certain criteria (personal preference, visual appeal, alignment with the prompt, etc.).

Each question contains two media samples and a question asking you to pick the best one according to certain criteria (personal preference, visual appeal, alignment with the prompt, etc.).

2025

Evaluate AI Assistant Responses

Scale AITextText Generation

Interact with an AI assistant and judge how good the assistant responses are. Where possible you will write better responses.

Interact with an AI assistant and judge how good the assistant responses are. Where possible you will write better responses.

2024

Education

Work History