Shehab Farag

Key Skills

Software

Data Annotation Tech

Labelbox

Top Subject Matter

No subject matter listed

Top Data Types

Audio

Text

Top Task Types

Emotion Recognition

Evaluation Rating

Prompt Response Writing SFT

RLHF

Freelancer Overview

I am an Arabic AI Data Specialist with hands-on experience for 6 months, in data labeling, annotation, and LLM evaluation, particularly within Reinforcement Learning from Human Feedback (RLHF) and Supervised Fine-Tuning (SFT) workflows. My background includes evaluating and ranking Arabic language model outputs for quality, factual accuracy, and safety, as well as annotating large-scale NLP datasets for sentiment analysis, named entity recognition, and intent classification. I have worked extensively with platforms like DataAnnotation.tech, Alignerr, and Labelbox, ensuring high inter-annotator consistency and contributing to the refinement of annotation guidelines. My medical education has strengthened my analytical reasoning and fact-verification skills, enabling me to deliver precise, reliable data for AI training and evaluation tasks across diverse domains.

Labeling Experience

Ai Lingustics Specialist

Data Annotation TechTextEvaluation Rating

1. Specific Data Tasks & Specializations Instruction Following & Constraint Adherence: Evaluated model performance based on complex multi-step instructions, ensuring 100% adherence to formatting, tone, and length constraints. Adversarial Testing (Red Teaming): Designed challenging "corner-case" prompts to trigger model hallucinations or safety violations, helping to refine the model's refusal boundaries. Arabic Linguistic Auditing: Specialized in rating Egyptian and Modern Standard Arabic (MSA) responses for naturalness, cultural relevance, and persona consistency. Fact-Checking & Research: Performed deep-dive verification of technical and medical claims using authoritative sources to ground model outputs in reality. 2. Project Size & Volume (Quantifiable Impact) Throughput: Managed a consistent volume of 30 complex rating tasks per week, maintaining high engagement with long-context windows . Scale: Contributed to large-scale datasets for leading A

1. Specific Data Tasks & Specializations Instruction Following & Constraint Adherence: Evaluated model performance based on complex multi-step instructions, ensuring 100% adherence to formatting, tone, and length constraints. Adversarial Testing (Red Teaming): Designed challenging "corner-case" prompts to trigger model hallucinations or safety violations, helping to refine the model's refusal boundaries. Arabic Linguistic Auditing: Specialized in rating Egyptian and Modern Standard Arabic (MSA) responses for naturalness, cultural relevance, and persona consistency. Fact-Checking & Research: Performed deep-dive verification of technical and medical claims using authoritative sources to ground model outputs in reality. 2. Project Size & Volume (Quantifiable Impact) Throughput: Managed a consistent volume of 30 complex rating tasks per week, maintaining high engagement with long-context windows . Scale: Contributed to large-scale datasets for leading A

2025

Education

Work History