Johnson Onuoha

Key Skills

Software

Mercor

Other

Top Subject Matter

No subject matter listed

Top Data Types

Video

Top Label Types

Segmentation

Action Recognition

Freelancer Overview

I am a detail-oriented AI Trainer and Data Annotation Specialist with hands-on experience in reviewing, labeling, and validating structured datasets to enhance AI model performance. My work has involved correcting text annotations, evaluating action-object relationships in both video and text data, and ensuring the accuracy of timestamps and segment boundaries. I am highly skilled at following detailed project guidelines, delivering high-quality, consistent results, and providing structured feedback to improve data quality. My technical proficiency includes using data annotation platforms, Microsoft Office, and Google Workspace, as well as tracking quality metrics through spreadsheets. I thrive in remote environments and am committed to delivering precise, reliable training data that supports the development of robust AI systems.

Labeling Experience

Atlas

MercorVideoSegmentationAction Recognition

Project Scope The project involved annotating first-person instructional videos showing how to change bicycle brake pads. The goal was to create accurate training data for AI systems to recognize human actions, tool usage, and object interactions during mechanical repair tasks. Data Labeling Tasks Divided videos into clear action-based segments Labeled main actions (e.g., loosening bolts, removing brake pads, installing new pads) Tagged related objects (e.g., Allen key, brake caliper, brake pads) Verified and corrected timestamps Ensured accurate action-object pairing and consistent terminology Project Size Multiple 5–15 minute videos 20–50 segments per video Hundreds of annotated action-object interactions Quality Measures Followed strict annotation guidelines Conducted detailed segment-by-segment reviews Maintained timestamp accuracy Ensured consistency and achieved high accuracy standards (95%+ benchmark)

Project Scope The project involved annotating first-person instructional videos showing how to change bicycle brake pads. The goal was to create accurate training data for AI systems to recognize human actions, tool usage, and object interactions during mechanical repair tasks. Data Labeling Tasks Divided videos into clear action-based segments Labeled main actions (e.g., loosening bolts, removing brake pads, installing new pads) Tagged related objects (e.g., Allen key, brake caliper, brake pads) Verified and corrected timestamps Ensured accurate action-object pairing and consistent terminology Project Size Multiple 5–15 minute videos 20–50 segments per video Hundreds of annotated action-object interactions Quality Measures Followed strict annotation guidelines Conducted detailed segment-by-segment reviews Maintained timestamp accuracy Ensured consistency and achieved high accuracy standards (95%+ benchmark)

2025 - 2025

Education

Work History