Mercy Olabode

Key Skills

Software

CloudFactory

Labelbox

Mindrift

Top Subject Matter

No subject matter listed

Top Data Types

Image

Text

Video

Top Task Types

Prompt Response Writing SFT

Text Generation

Text Summarization

Freelancer Overview

This project aimed to improve the fairness and accuracy of computer vision models by labeling diverse datasets representing real-world scenarios. I performed bounding box annotation and object tracking on images and videos using Keikaku, and conducted quality reviews through SRT. The goal was to ensure model performance across different demographics, environments, and conditions. The project involved thousands of media files, with strict quality standards requiring 95%+ accuracy. I followed detailed guidelines, ensured labeling consistency, and supported the delivery of high-quality training data for AI systems.

Labeling Experience

EGO-HOW-TO

MindriftVideoSegmentationAction Recognition

I worked on the EGO-HOW-TO project, labeling first-person videos to help AI models understand daily human activities. Using annotation tools, I applied bounding boxes, tracked hand-object interactions, and labeled actions like grabbing, cutting, or pouring. The project required frame-accurate tracking, attention to detail, and consistent application of complex activity labels. My work helped improve AI systems used in robotics, AR, and assistive tech.

I worked on the EGO-HOW-TO project, labeling first-person videos to help AI models understand daily human activities. Using annotation tools, I applied bounding boxes, tracked hand-object interactions, and labeled actions like grabbing, cutting, or pouring. The project required frame-accurate tracking, attention to detail, and consistent application of complex activity labels. My work helped improve AI systems used in robotics, AR, and assistive tech.

2024 - 2024

One vision/Fair perception

MindriftImageObject Detection

This project aimed to improve the fairness and accuracy of computer vision models by labeling diverse datasets representing real-world scenarios. I performed bounding box annotation and object tracking on images and videos and conducted quality reviews using SRT. The goal was to ensure model performance across different demographics, environments, and conditions. The project involved thousands of media files, with strict quality standards requiring 95%+ accuracy. I followed detailed guidelines, ensured labeling consistency, and supported the delivery of high-quality training data for AI systems.

This project aimed to improve the fairness and accuracy of computer vision models by labeling diverse datasets representing real-world scenarios. I performed bounding box annotation and object tracking on images and videos and conducted quality reviews using SRT. The goal was to ensure model performance across different demographics, environments, and conditions. The project involved thousands of media files, with strict quality standards requiring 95%+ accuracy. I followed detailed guidelines, ensured labeling consistency, and supported the delivery of high-quality training data for AI systems.

2024 - 2024

Large Language Model(Llama)

MindriftTextText Generation

Project: LLaMA – Large Language Model Annotation I worked on a large-scale LLaMA project aimed at improving the performance and reliability of a cutting-edge language model. My tasks included ranking AI-generated responses, rewriting text, and flagging inaccurate or harmful content. I followed detailed guidelines to assess outputs for truthfulness, clarity, relevance, and safety across a wide range of topics. The work required strong language skills, critical thinking, and close attention to detail. Quality standards were high, with a focus on producing training data that helped the model generate more accurate, helpful, and fair responses. This project sharpened my ability to evaluate nuanced language and contribute to the development of responsible AI systems.

Project: LLaMA – Large Language Model Annotation I worked on a large-scale LLaMA project aimed at improving the performance and reliability of a cutting-edge language model. My tasks included ranking AI-generated responses, rewriting text, and flagging inaccurate or harmful content. I followed detailed guidelines to assess outputs for truthfulness, clarity, relevance, and safety across a wide range of topics. The work required strong language skills, critical thinking, and close attention to detail. Quality standards were high, with a focus on producing training data that helped the model generate more accurate, helpful, and fair responses. This project sharpened my ability to evaluate nuanced language and contribute to the development of responsible AI systems.

2023 - 2024

Education

Work History