For employers

Hire this AI Trainer

Sign in or create an account to invite AI Trainers to your job.

LoginCreate account
Invite to Job
F
Fortunate Lebeya

Fortunate Lebeya

AI Data Annotation Specialist

South Africa flagJohannesburg, South Africa
$15.00/hrIntermediateOther

Key Skills

Software

Other

Top Subject Matter

AI Language Models
Reasoning Domain Expertise
Logic Domain Expertise

Top Data Types

TextText
AudioAudio
DocumentDocument

Top Task Types

RLHFRLHF
TranscriptionTranscription

Freelancer Overview

I am an experienced AI Trainer and Specialist with a deep focus on RLHF (Reinforcement Learning from Human Feedback), complex reasoning, and model safety. Since 2024, I have contributed to high-priority development projects at Outlier AI, Appen and RWS, where I was specifically selected for AI Safety initiatives. My work involves auditing model outputs for ethical compliance, bias mitigation, and harm prevention, as well as performing high-precision audio transcription to enhance multimodal speech-to-text accuracy. What sets me apart is the intersection of a BSc in Human Genetics and technical proficiency in Python (CS50). This unique combination allows me to evaluate model responses for scientific accuracy and logical consistency with a level of rigor that standard generalists cannot provide. A graduate of the McKinsey Forward and ALX programs, I am a highly disciplined remote professional dedicated to the meticulous, step-by-step reasoning required to build safe, reliable, and factually grounded frontier models.

IntermediateEnglish

Labeling Experience

AI Trainer / Contributor

OtherTextRLHF
As an AI Trainer/Contributor at Outlier, I contributed to Reinforcement Learning from Human Feedback (RLHF) projects by creating high-quality, original content. My main task involved writing, fact-checking, and optimizing responses to teach AI models improved logic, reasoning, and alignment. I maintained consistent quality by following detailed instructions and rigorous standards. • Crafted and evaluated text-based interactions to refine AI model capabilities. • Ensured style and content fidelity in prompt completion tasks. • Adhered to strict guidelines for factual accuracy and logical coherence. • Regularly reviewed AI-generated responses for rating and feedback.

As an AI Trainer/Contributor at Outlier, I contributed to Reinforcement Learning from Human Feedback (RLHF) projects by creating high-quality, original content. My main task involved writing, fact-checking, and optimizing responses to teach AI models improved logic, reasoning, and alignment. I maintained consistent quality by following detailed instructions and rigorous standards. • Crafted and evaluated text-based interactions to refine AI model capabilities. • Ensured style and content fidelity in prompt completion tasks. • Adhered to strict guidelines for factual accuracy and logical coherence. • Regularly reviewed AI-generated responses for rating and feedback.

2024 - Present

AI Specialist (Diamond Project)

OtherText
In the RWS Diamond Project, I served as an AI Specialist responsible for evaluating system responses for accuracy and instruction alignment. My work included extensive review of complex, multi-turn AI-generated outputs requiring subject matter expertise. I collaborated on high-priority tasks emphasizing pinpoint accuracy and nuance. • Provided detailed feedback on AI outputs for safety and factual correctness. • Participated in quality assurance for LLM instruction following. • Specialized in multi-step, context-driven evaluation. • Applied advanced linguistic and scientific analysis for improvement.

In the RWS Diamond Project, I served as an AI Specialist responsible for evaluating system responses for accuracy and instruction alignment. My work included extensive review of complex, multi-turn AI-generated outputs requiring subject matter expertise. I collaborated on high-priority tasks emphasizing pinpoint accuracy and nuance. • Provided detailed feedback on AI outputs for safety and factual correctness. • Participated in quality assurance for LLM instruction following. • Specialized in multi-step, context-driven evaluation. • Applied advanced linguistic and scientific analysis for improvement.

2026 - 2026

AI safety project

TextPrompt Response Writing SFT
The Pangolin Safety Project is a high-priority AI alignment initiative focused on adversarial red-teaming to refine the ethical boundaries of frontier LLMs. As a contributor, I performed complex data labeling tasks including adversarial prompting to test model "jailbreaks," safety categorization of policy violations (such as PII, bias, and dangerous content), and response rewriting to balance helpfulness with strict harm mitigation. Working within a large-scale global expert operation, I adhered to rigorous quality measures such as providing "Chain-of-Thought" rationales for every label and maintaining 95%+ accuracy against gold-standard benchmarks to ensure the logical consistency and safety of public-facing models.

The Pangolin Safety Project is a high-priority AI alignment initiative focused on adversarial red-teaming to refine the ethical boundaries of frontier LLMs. As a contributor, I performed complex data labeling tasks including adversarial prompting to test model "jailbreaks," safety categorization of policy violations (such as PII, bias, and dangerous content), and response rewriting to balance helpfulness with strict harm mitigation. Working within a large-scale global expert operation, I adhered to rigorous quality measures such as providing "Chain-of-Thought" rationales for every label and maintaining 95%+ accuracy against gold-standard benchmarks to ensure the logical consistency and safety of public-facing models.

2024 - 2025

AI Data Annotation Specialist / Audio Transcription

OtherAudioTranscription
I performed high-precision audio transcription and linguistic analysis to enhance speech-to-text AI model accuracy as part of data annotation initiatives. The role required attention to detail and scientific language proficiency to generate clean transcriptions. My contributions directly supported training datasets for speech recognition AI. • Transcribed audio with high fidelity to text format. • Identified and corrected linguistic errors or ambiguities. • Improved dataset quality for speech-to-text models. • Maintained accuracy in scientific terminology and subject matter.

I performed high-precision audio transcription and linguistic analysis to enhance speech-to-text AI model accuracy as part of data annotation initiatives. The role required attention to detail and scientific language proficiency to generate clean transcriptions. My contributions directly supported training datasets for speech recognition AI. • Transcribed audio with high fidelity to text format. • Identified and corrected linguistic errors or ambiguities. • Improved dataset quality for speech-to-text models. • Maintained accuracy in scientific terminology and subject matter.

Not specified

Education

E

edX

CS50's Introduction to Programming with Python, Computer Science and Software Development

CS50's Introduction to Programming with Python
2026 - 2026
M

McKinsey and Company Forward Program

Forward Program, Business Leadership & Digital Innovation, Professional Development & Strategic Problem Solving, Digital Transformation

Forward Program
2024 - 2024

Work History

L

Linksfield Hospital

Student Medical Technician

Johannesburg
2021 - 2023
L

Linksfield Hospital

Admin Clerk

Johannesburg
2021 - 2023