Abdulkareem A.

Key Skills

Software

Labelbox

Micro1

Top Subject Matter

No subject matter listed

Top Data Types

Video

Image

Text

Top Task Types

Bounding Box

Segmentation

Classification

Object Detection

Text Generation

RLHF

Data Collection

Polygon

Freelancer Overview

I've been doing data labeling and AI training work for about a year across different platforms and companies. Most of my experience has been with video annotation — bounding boxes, object tagging, action recognition — but I've also worked extensively on text, audio, and image datasets. At Micro1 I handled around 300 assets a month across multimodal projects and maintained a 90%+ accuracy rate consistently. I also did a POV video collection project there where I generated over 100 hours of egocentric training footage for perception model training, finishing with a 95% acceptance rate. What I think sets me apart is the combination of volume, consistency, and attention to detail. I don't just complete tasks — I actually read the guidelines carefully, flag edge cases, and care about whether the data is genuinely useful for the model being trained. I also have a geology degree, which means I'm comfortable working with technical, structured data and following complex analytical frameworks. That background has made me a stronger annotator, especially on scientific or environmental datasets.

Labeling Experience

AI training specialist

VideoSegmentation

This project involved annotating video datasets used to train and improve machine learning models. Tasks included object identification, action recognition, scene classification, and contextual labeling within video frames — all following detailed annotation guidelines provided for each project cycle. In terms of scale, I processed over 300 video assets per month as part of a distributed remote annotation team. To maintain quality, I followed strict labeling guidelines on every task, carried out peer-level QA reviews to catch inconsistencies before submission, and held a 90%+ accuracy rate consistently across all submissions.

2026 - 2026

Education