Peter Nganga

Key Skills

Software

Appen

Clickworker

CloudFactory

CVAT

Labelbox

Lionbridge

Mindrift

Roboflow

Toloka

Top Subject Matter

Language Models & Text Generation

Multimedia Annotation

Autonomous Vehicles & Computer Vision

Top Data Types

Audio

Image

Text

Top Task Types

Audio Recording

Evaluation Rating

Prompt Response Writing SFT

Text Generation

Translation Localization

Freelancer Overview

I am an experienced AI Data Annotator and Computer Vision Specialist with 5+ years delivering high-quality image, video, and text datasets for production ML systems. I’ve worked with global vendors (Appen, Toloka/Mindrift, Lionbridge, RWS Group) on projects spanning face data, product identification, sports action recognition, and medical datasets. I am highly proficient with Roboflow, Labelbox, CVAT, and Supervisely — from dataset ingestion and augmentation to export-ready formats for training. I consistently deliver >95%–98% annotation accuracy, build QA workflows, and create annotation guidelines that improve model performance and team throughput. Why I stand out Direct, hands-on experience preparing and managing Roboflow datasets and pipelines (dataset import, class mapping, augmentation presets, export for YOLO/COCO/TF). Deep expertise in image annotation tasks: bounding boxes, polygons, segmentation masks, keypoints, image classification and object tracking in video. Proven track record improving label quality and model performance through QA processes and cross-review. Strong collaboration with ML engineers and data scientists to ensure datasets are production-ready.

Labeling Experience

Social Media / Product Dataset

RoboflowImageBounding BoxPolygon

I contributed to a large-scale image data labeling project focused on social media and product datasets designed for computer vision model training. Tasks included: Creating and managing Roboflow datasets (importing images, defining classes, dataset augmentation, and exporting in YOLO/COCO formats). Performing bounding box, polygon, segmentation, and keypoint annotation to identify faces, objects, and brand elements. Ensuring label accuracy above 98% through strict QA checks and cross-review processes. Collaborating with machine learning engineers to optimize dataset structure for object detection and classification models. The project involved labeling over 20,000 images across multiple categories and directly contributed to improving AI model precision in visual recognition and tagging tasks.

I contributed to a large-scale image data labeling project focused on social media and product datasets designed for computer vision model training. Tasks included: Creating and managing Roboflow datasets (importing images, defining classes, dataset augmentation, and exporting in YOLO/COCO formats). Performing bounding box, polygon, segmentation, and keypoint annotation to identify faces, objects, and brand elements. Ensuring label accuracy above 98% through strict QA checks and cross-review processes. Collaborating with machine learning engineers to optimize dataset structure for object detection and classification models. The project involved labeling over 20,000 images across multiple categories and directly contributed to improving AI model precision in visual recognition and tagging tasks.

2024 - 2024

Medical Data Annotation Specialist

MindriftTextEntity Ner ClassificationSegmentation

Conducted data extraction and preprocessing for AI-driven healthcare research. Assisted in organizing and structuring large datasets by identifying and labeling critical patterns in patient records. Worked with machine learning teams to refine datasets used in predictive analytics, ensuring data accuracy and consistency. Utilized annotation tools to classify medical information and support model training for event prediction in clinical settings.

Conducted data extraction and preprocessing for AI-driven healthcare research. Assisted in organizing and structuring large datasets by identifying and labeling critical patterns in patient records. Worked with machine learning teams to refine datasets used in predictive analytics, ensuring data accuracy and consistency. Utilized annotation tools to classify medical information and support model training for event prediction in clinical settings.

2023 - 2024

Sports Video Data Labeling for Action Recognition and Object Tracking

CVATVideoBounding BoxSegmentation

This project involved labeling video data from football matches to train machine learning models in action recognition and player tracking. The tasks included drawing bounding boxes around players, tracking their movement across frames, and identifying key events such as passes, tackles, and goals. The project required a meticulous approach to ensure high precision in annotations, particularly given the fast pace of gameplay. I worked on a dataset containing 5 hours of match footage, where I applied bounding box techniques to identify players and other objects, as well as action labeling to categorize specific in-game actions. Quality measures included cross-verification of annotations to maintain a high standard of accuracy and consistency throughout the dataset.

This project involved labeling video data from football matches to train machine learning models in action recognition and player tracking. The tasks included drawing bounding boxes around players, tracking their movement across frames, and identifying key events such as passes, tackles, and goals. The project required a meticulous approach to ensure high precision in annotations, particularly given the fast pace of gameplay. I worked on a dataset containing 5 hours of match footage, where I applied bounding box techniques to identify players and other objects, as well as action labeling to categorize specific in-game actions. Quality measures included cross-verification of annotations to maintain a high standard of accuracy and consistency throughout the dataset.

2022 - 2023

Multilingual Conversational AI Data Annotation and Evaluation

LabelboxTextText GenerationEvaluation Rating

I worked on multilingual conversational AI data annotation and evaluation projects aimed at training and refining language models. The scope involved labeling text and audio data to train AI models for accurate conversation generation in both English and Kiswahili. Specific tasks included generating conversation samples, labeling intents and responses, and evaluating AI-generated outputs for relevance, clarity, and factual accuracy. My contribution directly improved the model's conversational accuracy by 20%, and I developed documentation and feedback mechanisms that increased overall model efficiency by 15%. The project focused on enhancing the AI's ability to handle multilingual content, understand context, and improve empathy in responses. Collaboration with cross-functional teams was crucial in refining the AI models and ensuring consistent, high-quality annotations.

I worked on multilingual conversational AI data annotation and evaluation projects aimed at training and refining language models. The scope involved labeling text and audio data to train AI models for accurate conversation generation in both English and Kiswahili. Specific tasks included generating conversation samples, labeling intents and responses, and evaluating AI-generated outputs for relevance, clarity, and factual accuracy. My contribution directly improved the model's conversational accuracy by 20%, and I developed documentation and feedback mechanisms that increased overall model efficiency by 15%. The project focused on enhancing the AI's ability to handle multilingual content, understand context, and improve empathy in responses. Collaboration with cross-functional teams was crucial in refining the AI models and ensuring consistent, high-quality annotations.

2020 - 2021

Education

Work History