Moses Midega

Key Skills

Software

CVAT

Top Subject Matter

No subject matter listed

Top Data Types

Image

Top Task Types

Bounding Box

Entity Ner Classification

Segmentation

Freelancer Overview

I am a detail-oriented data annotation specialist with over 3 years of professional experience, transitioning from risk mitigation and audit compliance roles into AI training data and machine learning support. My background has equipped me with a strong analytical mindset and exceptional attention to detail, which I now apply to high-quality data labeling projects involving image, video, and text annotation. I am skilled in using annotation platforms like Labelbox and CVAT, and have hands-on experience with bounding box, polygon, semantic segmentation, and text classification tasks. My work in business development and risk assessment has sharpened my ability to understand project context, maintain data integrity, and ensure accurate dataset preparation for AI models. I am fluent in both English and Swahili, comfortable working in collaborative environments, and committed to delivering reliable, precisely labeled datasets that drive data-driven decision-making.

Labeling Experience

Precision Annotation for 2D Object Detection in Urban Driving Scenes

CVATImageBounding BoxEntity Ner Classification

Scope & Tasks: Led the annotation for a key dataset to improve an autonomous vehicle perception model's performance in dense urban environments. The primary task was 2D bounding box annotation with precise classification on high-resolution camera images. Specific Tasks Performed: Accurately drew bounding boxes around critical dynamic objects (cars, trucks, motorcycles, bicycles, pedestrians). Classified objects into fine-grained categories (e.g., "pedestrian-walking," "pedestrian-with-bicycle," "construction-vehicle"). Tagged occlusion levels (fully visible, partially occluded, heavily occluded) and truncation status for each object. Collaborated with project leads to resolve edge cases and ensure label consistency across a team of 5 annotators. Project Size: Annotated and reviewed over 15,000 high-resolution images, resulting in ~220,000 labeled object instances. Quality Measures: Adhered to a detailed, client-provided annotation guideline (50+ pages). Achieved and maintained

Scope & Tasks: Led the annotation for a key dataset to improve an autonomous vehicle perception model's performance in dense urban environments. The primary task was 2D bounding box annotation with precise classification on high-resolution camera images. Specific Tasks Performed: Accurately drew bounding boxes around critical dynamic objects (cars, trucks, motorcycles, bicycles, pedestrians). Classified objects into fine-grained categories (e.g., "pedestrian-walking," "pedestrian-with-bicycle," "construction-vehicle"). Tagged occlusion levels (fully visible, partially occluded, heavily occluded) and truncation status for each object. Collaborated with project leads to resolve edge cases and ensure label consistency across a team of 5 annotators. Project Size: Annotated and reviewed over 15,000 high-resolution images, resulting in ~220,000 labeled object instances. Quality Measures: Adhered to a detailed, client-provided annotation guideline (50+ pages). Achieved and maintained

2023 - 2023

Education

Work History