Kenneth Owuor

Key Skills

Software

CVAT

Label Studio

LabelImg

Top Subject Matter

No subject matter listed

Top Data Types

Image

Text

Video

Top Task Types

Bounding Box

Freelancer Overview

I am an experienced image annotation specialist with a strong background in creating training data for AI applications such as facial recognition, identity matching, and accessory removal. I have contributed to high-precision datasets through my work with Moonvalley and HumanSignal, where I followed structured protocols to ensure consistent, high-quality annotations that support robust machine learning models. I specialize in tagging occlusions, matching identities across frames, and refining visual datasets for biometric and surveillance applications. My methodical approach, combined with a deep understanding of annotation quality assurance workflows, enables me to deliver reliable and reproducible results. This ultimately enhances model performance and reduces bias.

Labeling Experience

moonvalley image annotation

LabelimgImageBounding Box

The Moonvalley AI Platform supports image annotation workflows that contribute directly to high-quality video generation. As an image annotation specialist, I’ve worked on tasks that involve tagging facial features, matching identities across frames, and removing accessories such as glasses or masks to ensure clean training data for Marey, Moonvalley’s proprietary video model. These annotations help the model learn realistic motion, lighting, and physics by isolating core visual elements and minimizing occlusions. My work ensures that the AI-generated outputs maintain artistic integrity, legal safety, and production-grade quality—critical for professional filmmaking and creative control.

The Moonvalley AI Platform supports image annotation workflows that contribute directly to high-quality video generation. As an image annotation specialist, I’ve worked on tasks that involve tagging facial features, matching identities across frames, and removing accessories such as glasses or masks to ensure clean training data for Marey, Moonvalley’s proprietary video model. These annotations help the model learn realistic motion, lighting, and physics by isolating core visual elements and minimizing occlusions. My work ensures that the AI-generated outputs maintain artistic integrity, legal safety, and production-grade quality—critical for professional filmmaking and creative control.

2024 - 2025

identity matching

Label StudioImage

Identity matching is a data annotation project that focuses on determining whether different records refer to the same real-world person. The scope of the project includes comparing personal data such as names, photos, ID numbers, and contact details to detect and merge duplicate identities, verify user information, and support fraud detection systems across industries like banking, telecommunications, healthcare, and government services. Human annotators perform several labeling tasks such as pairwise matching, where two records are compared and labeled as “match” or “no match,” triplet comparisons, attribute verification, image-to-text matching, and grouping multiple profiles that belong to one individual. The project size varies depending on client needs, ranging from small pilot projects with tens of thousands of record pairs to large-scale enterprise projects involving millions of comparisons. To ensure high accuracy and reliability, strict quality measures are followed, including

Identity matching is a data annotation project that focuses on determining whether different records refer to the same real-world person. The scope of the project includes comparing personal data such as names, photos, ID numbers, and contact details to detect and merge duplicate identities, verify user information, and support fraud detection systems across industries like banking, telecommunications, healthcare, and government services. Human annotators perform several labeling tasks such as pairwise matching, where two records are compared and labeled as “match” or “no match,” triplet comparisons, attribute verification, image-to-text matching, and grouping multiple profiles that belong to one individual. The project size varies depending on client needs, ranging from small pilot projects with tens of thousands of record pairs to large-scale enterprise projects involving millions of comparisons. To ensure high accuracy and reliability, strict quality measures are followed, including

2025

Education

Work History