Kevin R.

Key Skills

Software

Other

Scale AI

Top Subject Matter

LLM safety and language model training (legal, medical, general domain)

Speech recognition training data QA (multi-speaker accents, noisy audio)

Multi-modal AI/ML training dataset annotation (legal, medical, general)

Top Data Types

Text

Audio

Image

Document

Top Task Types

Transcription

Classification

Red Teaming

Freelancer Overview

Senior Data Labeler & AI Trainer, Global AI Data Services (Text/LLM RLHF). Core strengths include Other and Scale AI. Education includes Bachelor's Degree Program, University of Nairobi. AI-training focus includes data types such as Text, Audio, and Computer Code and labeling workflows including Evaluation, Rating, and Transcription.

Labeling Experience

Senior Data Labeler & AI Trainer (LLM safety evaluation)

OtherTextRed Teaming

Evaluated 10,000+ model responses for helpfulness, harmlessness, and honesty to support LLM safety improvements. Applied structured criteria to identify quality and safety issues in generated text. Contributed to reinforcement learning from human feedback prompt evaluation activities. • Rated responses using safety and helpfulness dimensions. • Identified failure modes affecting honesty and safety. • Supported RLHF prompt evaluation workflows. • Helped teams refine datasets for safer LLM outputs.

2022 - Present

Technical projects supporting labeling and AI training (workflow, QA, bilingual curation)

Other

Developed a semi-automated annotation workflow using Python scripts and labeling tool APIs to reduce manual labeling time. Designed QA checklists and an error taxonomy to improve output consistency across AI training datasets. Curated and cleaned bilingual datasets (English/Swahili) to support NLP training for low-resource language tasks. • Implemented workflow automation to cut manual labeling by 30%. • Created scalable QA checklist and error taxonomy for 25-member team. • Curated bilingual datasets to improve NLP model performance. • Supported dataset consistency through structured QA processes.

2022 - Present

Senior Data Labeler & AI Trainer, Global AI Data Services

Scale AITextClassification

Annotated and labeled 500,000+ data points across text, image, and audio modalities for major AI/ML training projects. Maintained 99%+ annotation accuracy using strict labeling guidelines across legal, medical, and general-domain datasets. Reviewed and validated junior labeler outputs to reduce error rates through structured feedback cycles. • Labeled large volumes of training data for AI/ML model development. • Applied rigorous guidelines to sustain 99%+ accuracy. • Reduced error rates by 18% via feedback cycles. • Delivered work ahead of schedule using labeling platform workflows.

2022 - Present

Senior Data Labeler & AI Trainer, Global AI Data Services (Text/LLM RLHF)

OtherText

Performed RLHF evaluations to improve large language model output quality and safety alignment. Labeled and annotated training data to support major AI/ML model projects across language understanding tasks. Applied quality standards to maintain consistent dataset behavior. • Conducted helpfulness/harmlessness/honesty evaluations for LLM safety improvements. • Supported reinforcement learning from human feedback workflows. • Maintained accuracy targets for labeled text datasets. • Provided feedback loops tied to error reduction outcomes.

2022 - Present

Quality Assurance lead - Transcription & Annotation, Freelance & Contract Projects

OtherAudioTranscription

Audited and refined transcription and annotation files produced by junior annotators for AI training datasets. Specialized in multi-speaker audio with heavy accents and noisy environments to meet high clarity standards. Delivered clean, labeled audio suitable for speech recognition model training. • Audited transcription outputs for accuracy and consistency. • Produced labeled audio data totaling 200+ hours for training. • Developed QA checklists and standardized processes across 10+ annotators. • Ensured high-quality labeled outputs despite challenging audio conditions.

2020 - 2022

Education

Work History