Okeowo Micheal

Key Skills

Software

Labelbox

Label Studio

Other

Internal/Proprietary Tooling

CVAT

Mercor

Top Subject Matter

AI Model Evaluation

Video and Audio Content

Rubric Writing

Top Data Types

Video

Text

Document

Top Task Types

Action Recognition

Bounding Box

Classification

Evaluation/Rating

Prompt + Response Writing (SFT)

Freelancer Overview

AI evaluation expert with expertise in text, video, audio, and document annotation. Skilled in enhancing dataset quality, refining generative AI outputs, and optimizing annotation workflows to improve model reliability and fairness.

Labeling Experience

Generalist Video Annotator & AI Trainer

Label StudioVideoClassification

As a Generalist Video Annotator and AI Trainer at Mercor USA, I maintained guideline compliance while annotating video and audio data, rewriting rubrics, and developing model-stumping prompts. I collaborated cross-functionally to resolve annotation and labeling issues, ensuring consistency in adversarial prompt workflows and adapting to evolving guidelines. My responsibilities included quality assurance, rubric standardization, and workflow improvements to optimize annotation processes. • Maintained 98% guideline adherence, completing 49 video annotation tasks weekly with SRT and Parimango. • Annotated and transcribed audio datasets, delivering over 1,800 entries and reducing final-stage errors by 17%. • Rewrote 300+ evaluation rubrics, improving clarity and standardization across model evaluations. • Created, tested, and refined 200+ adversarial prompts, increasing identification of LLM edge-case failures.

As a Generalist Video Annotator and AI Trainer at Mercor USA, I maintained guideline compliance while annotating video and audio data, rewriting rubrics, and developing model-stumping prompts. I collaborated cross-functionally to resolve annotation and labeling issues, ensuring consistency in adversarial prompt workflows and adapting to evolving guidelines. My responsibilities included quality assurance, rubric standardization, and workflow improvements to optimize annotation processes. • Maintained 98% guideline adherence, completing 49 video annotation tasks weekly with SRT and Parimango. • Annotated and transcribed audio datasets, delivering over 1,800 entries and reducing final-stage errors by 17%. • Rewrote 300+ evaluation rubrics, improving clarity and standardization across model evaluations. • Created, tested, and refined 200+ adversarial prompts, increasing identification of LLM edge-case failures.

2025 - Present

Audio Annotation Expert

AudioTranscription

- Annotated and transcribed spoken content from video datasets into structured text aligned with predefined labeling guidelines for AI training use. - Conducted QA review of audio annotations prior to delivery, correcting inconsistencies and improving transcription accuracy across multimodal datasets. - Validated alignment between audio content and annotations to ensure semantic correctness and consistency across labeled training data. - Maintained standardized review practices to reduce transcription errors and improve the reliability of audio datasets used in model training.

- Annotated and transcribed spoken content from video datasets into structured text aligned with predefined labeling guidelines for AI training use. - Conducted QA review of audio annotations prior to delivery, correcting inconsistencies and improving transcription accuracy across multimodal datasets. - Validated alignment between audio content and annotations to ensure semantic correctness and consistency across labeled training data. - Maintained standardized review practices to reduce transcription errors and improve the reliability of audio datasets used in model training.

2026 - 2026

Model Stumper

TextPrompt Response Writing SFT

- Designed adversarial prompts to test AI model reasoning under structured constraints and evaluate failure cases in LLM behavior. - Generated over 200 structured prompts targeting edge-case reasoning, instruction-following, and contextual understanding failures. - Analyzed model responses to identify weaknesses in reasoning accuracy, consistency, and instruction adherence. - Iteratively refined prompt strategies to improve the effectiveness of evaluation tests for LLM robustness and behavioral analysis.

- Designed adversarial prompts to test AI model reasoning under structured constraints and evaluate failure cases in LLM behavior. - Generated over 200 structured prompts targeting edge-case reasoning, instruction-following, and contextual understanding failures. - Analyzed model responses to identify weaknesses in reasoning accuracy, consistency, and instruction adherence. - Iteratively refined prompt strategies to improve the effectiveness of evaluation tests for LLM robustness and behavioral analysis.

2026 - 2026

Generalist Video Annotator

VideoSegmentation

- Annotated large-scale video datasets involving actions, objects, and scene transitions to support computer vision and multimodal AI model training. - Applied structured labeling across SRT and Parimango platforms, ensuring consistency in frame-level video annotation outputs aligned with evolving project guidelines. - Performed quality assurance checks on annotated video data before submission, maintaining a quality score above 4.0/5.0 and 98% guideline adherence. - Reviewed and corrected video annotations across multiple datasets to ensure accuracy, consistency, and alignment with training requirements for AI systems.

- Annotated large-scale video datasets involving actions, objects, and scene transitions to support computer vision and multimodal AI model training. - Applied structured labeling across SRT and Parimango platforms, ensuring consistency in frame-level video annotation outputs aligned with evolving project guidelines. - Performed quality assurance checks on annotated video data before submission, maintaining a quality score above 4.0/5.0 and 98% guideline adherence. - Reviewed and corrected video annotations across multiple datasets to ensure accuracy, consistency, and alignment with training requirements for AI systems.

2025 - 2026

Generalist AI Trainer

ImageBounding Box

- Annotated image-based document datasets by drawing bounding boxes around questions and corresponding answers, ensuring accurate spatial alignment between paired text elements within structured layouts. - Worked with document images containing embedded question–answer pairs, identifying and labeling both components only when full contextual elements were present for validation. - Applied strict annotation rules requiring the complete presence of question text, answer text, and image context before bounding box creation to ensure dataset integrity and labeling consistency. - Performed quality checks on bounding box annotations to ensure correct pairing of questions and answers, reducing labeling inconsistencies and improving dataset reliability for downstream model training.

- Annotated image-based document datasets by drawing bounding boxes around questions and corresponding answers, ensuring accurate spatial alignment between paired text elements within structured layouts. - Worked with document images containing embedded question–answer pairs, identifying and labeling both components only when full contextual elements were present for validation. - Applied strict annotation rules requiring the complete presence of question text, answer text, and image context before bounding box creation to ensure dataset integrity and labeling consistency. - Performed quality checks on bounding box annotations to ensure correct pairing of questions and answers, reducing labeling inconsistencies and improving dataset reliability for downstream model training.

2025 - 2025

Show more

Education

Work History