Samuel Manyeki

Key Skills

Software

Appen

Data Annotation Tech

Figure Eight

Labelbox

Label Studio

Remotasks

Toloka

Telus

Other

Internal/Proprietary Tooling

Scale AI

Top Subject Matter

No subject matter listed

Top Data Types

Image

Text

Video

Top Label Types

Bounding Box

Classification

Evaluation Rating

Prompt Response Writing SFT

Translation Localization

Freelancer Overview

Profile Overview Multilingual AI Data Trainer and Evaluator with hands-on experience in annotation, data labeling, model evaluation, and prompt engineering for global AI projects. Contributed to RWS TrainAI and Meta initiatives, including the Diamond Faceswap (Ruby & Parimango) image annotation project, where I specialized in data quality, classification, and cultural relevance for multilingual datasets. Combining technical accuracy with linguistic precision in English (C2 UK/C1 US) and Swahili (C1), I bring a unique ability to enhance AI systems through nuanced data interpretation and quality control. Skilled in Labelbox, Label Studio, Power BI, and Tableau, I focus on delivering clean, reliable datasets that improve AI model performance. Currently pursuing a BSc in Computer Science (online), I continuously expand my expertise in data systems, machine learning, and applied AI solutions.

Labeling Experience

Project 3: DataAnnotation.Tech - Multilingual Text Annotation & Entity Recognition Projects

LabelboxTextEntity Ner ClassificationText Generation

Engaged in independent and platform-based practice projects to refine entity recognition, translation/localization, and evaluation/rating of multilingual text data (English–Swahili). Focused on labeling consistency, linguistic nuance, and cultural accuracy to improve natural language processing datasets. Utilized Labelbox to structure annotation workflows and perform data quality checks for model fine-tuning.

Engaged in independent and platform-based practice projects to refine entity recognition, translation/localization, and evaluation/rating of multilingual text data (English–Swahili). Focused on labeling consistency, linguistic nuance, and cultural accuracy to improve natural language processing datasets. Utilized Labelbox to structure annotation workflows and perform data quality checks for model fine-tuning.

2024

Project 2: Meta AI - Model Evaluation & Prompt Engineering Support

Label StudioTextClassificationText Generation

Executed model evaluation and content review tasks for AI language models, assessing accuracy, bias, and factuality of generated outputs. Applied prompt + response testing (SFT) for instruction-tuning datasets and participated in translation/localization tasks between English and Swahili to enhance multilingual accuracy. Supported content safety flagging workflows and QA reviews to improve linguistic precision.

Executed model evaluation and content review tasks for AI language models, assessing accuracy, bias, and factuality of generated outputs. Applied prompt + response testing (SFT) for instruction-tuning datasets and participated in translation/localization tasks between English and Swahili to enhance multilingual accuracy. Supported content safety flagging workflows and QA reviews to improve linguistic precision.

2023

Project 1: RWS TrainAI - Diamond Faceswap Annotation – Ruby & Parimango Phases

Label StudioImageBounding BoxSegmentation

Contributed to large-scale AI image annotation for facial recognition dataset training under RWS TrainAI (Meta Project). Applied bounding box and segmentation techniques to identify and classify facial attributes with precision. Conducted evaluation/rating tasks to validate labeling consistency and guideline compliance. Maintained annotation accuracy above 98% while collaborating on refining task protocols for improved model efficiency.

Contributed to large-scale AI image annotation for facial recognition dataset training under RWS TrainAI (Meta Project). Applied bounding box and segmentation techniques to identify and classify facial attributes with precision. Conducted evaluation/rating tasks to validate labeling consistency and guideline compliance. Maintained annotation accuracy above 98% while collaborating on refining task protocols for improved model efficiency.

2022

Education

Work History