Stephen Kamau

Key Skills

Software

CVAT

Top Subject Matter

No subject matter listed

Top Data Types

Image

Top Label Types

Bounding Box

Polygon

Segmentation

Classification

Fine Tuning

Freelancer Overview

I am an AI/ML engineer with over three years of hands-on experience building and deploying machine learning solutions, with a strong emphasis on high-quality data labeling and AI training data pipelines. My expertise spans computer vision, NLP, and LLM-based systems, where I have designed scalable workflows for extracting, processing, and annotating large datasets—including medical imaging, legal and financial documents, and e-commerce text. I have implemented OCR and object detection systems using tools such as YOLO, Mask R-CNN, and Tesseract OCR, and fine-tuned transformer-based models like BERT and RoBERTa to improve model accuracy and domain adaptation. What sets me apart is my strong focus on data quality, scalability, and reproducibility. I have built end-to-end data pipelines using Airflow, DBT, and Snowflake to ensure reliable data validation, versioning, and monitoring in production environments. My work integrates MLOps best practices to maintain clean, well-structured training datasets that support accurate, fair, and robust AI systems across healthcare, finance, and logistics domains.

Labeling Experience

Medical & Document AI Data Annotation for Computer Vision and NLP Systems

CVATImageBounding BoxPolygon

Led end-to-end data annotation workflows for computer vision and NLP models across healthcare, finance, and legal domains. Annotated and validated medical imaging datasets using bounding boxes, polygon segmentation, and diagnostic classification to support object detection and disease identification models (YOLO, Mask R-CNN). Designed quality control protocols to ensure high inter-annotator agreement and dataset consistency for production-grade AI systems.

Led end-to-end data annotation workflows for computer vision and NLP models across healthcare, finance, and legal domains. Annotated and validated medical imaging datasets using bounding boxes, polygon segmentation, and diagnostic classification to support object detection and disease identification models (YOLO, Mask R-CNN). Designed quality control protocols to ensure high inter-annotator agreement and dataset consistency for production-grade AI systems.

2022 - 2022

Education

Work History