Mazza Ibrahim

Key Skills

Software

Labelbox

Remotasks

Scale AI

Top Subject Matter

No subject matter listed

Top Data Types

Audio

Image

Text

Top Task Types

Classification

Emotion Recognition

Prompt Response Writing SFT

Translation Localization

Freelancer Overview

I am a native Arabic speaker with over 3 years of hands-on experience in data labeling and linguistic annotation for AI systems. I’ve contributed to projects involving image tagging, intent classification, sentiment annotation, named entity recognition (NER), and transcription validation. My work has supported the development of multilingual LLMs and AI products requiring high-quality Arabic datasets, including Modern Standard Arabic and regional dialects. I have worked on platforms such as Remotasks and labelbox, where I followed detailed guidelines to complete tasks with accuracy and consistency. I am experienced in QA review processes, long-form transcription alignment, and content quality evaluation. My strengths lie in linguistic sensitivity, adherence to annotation protocols, and delivering reliable, structured data for training advanced AI systems.

Labeling Experience

Arabic LLM Prompt Evaluation & Text Labeling Project

LabelboxTextText GenerationPrompt Response Writing SFT

Worked on large-scale Arabic language data labeling projects for LLM development, focusing on text quality assessment, intent classification, prompt/response pair evaluations, and NER tasks. I evaluated generated text outputs for fluency, correctness, and cultural alignment, helping to improve the accuracy of AI systems trained in Arabic. The project included daily task batches exceeding 1,000 items, with stringent QA protocols and alignment to linguistic standards across Modern Standard Arabic and Gulf/North African dialects. I consistently maintained high annotation accuracy and met fast turnaround times, contributing to model fine-tuning for multilingual NLP.

Worked on large-scale Arabic language data labeling projects for LLM development, focusing on text quality assessment, intent classification, prompt/response pair evaluations, and NER tasks. I evaluated generated text outputs for fluency, correctness, and cultural alignment, helping to improve the accuracy of AI systems trained in Arabic. The project included daily task batches exceeding 1,000 items, with stringent QA protocols and alignment to linguistic standards across Modern Standard Arabic and Gulf/North African dialects. I consistently maintained high annotation accuracy and met fast turnaround times, contributing to model fine-tuning for multilingual NLP.

2024 - 2024

Arabic Audio Transcription Review & QA

LabelboxAudioTranslation LocalizationAudio Recording

Performed quality assurance for Arabic audio transcription projects by reviewing and correcting machine-generated transcriptions for accuracy, fluency, and alignment with audio. Tasks involved identifying transcription errors, adjusting punctuation, verifying dialect-specific terms, and ensuring time-aligned segments matched spoken content. Covered various dialects including Gulf Arabic, Levantine, and Modern Standard Arabic. Worked on datasets used to train Arabic voice assistants and improve ASR (Automatic Speech Recognition) systems. Maintained a QA accuracy rate above 97% while adhering to detailed linguistic and formatting guidelines.

Performed quality assurance for Arabic audio transcription projects by reviewing and correcting machine-generated transcriptions for accuracy, fluency, and alignment with audio. Tasks involved identifying transcription errors, adjusting punctuation, verifying dialect-specific terms, and ensuring time-aligned segments matched spoken content. Covered various dialects including Gulf Arabic, Levantine, and Modern Standard Arabic. Worked on datasets used to train Arabic voice assistants and improve ASR (Automatic Speech Recognition) systems. Maintained a QA accuracy rate above 97% while adhering to detailed linguistic and formatting guidelines.

2024

Arabic Image Annotation for OCR and Object Detection

LabelboxImageBounding Box

Contributed to high-volume Arabic image annotation tasks aimed at training computer vision models for OCR and object detection in Arabic-script environments. Tasks included drawing precise bounding boxes around handwritten and printed Arabic text, labeling contextual image regions (e.g., storefront signs, license plates, packaging), and tagging image quality for training robust OCR models. Ensured accurate transcription of Arabic content, including dialectal variations, diacritics, and ligature forms. Maintained over 95% QA compliance while processing over 500+ images per batch, helping develop region-aware AI tools for visual recognition in MENA markets.

Contributed to high-volume Arabic image annotation tasks aimed at training computer vision models for OCR and object detection in Arabic-script environments. Tasks included drawing precise bounding boxes around handwritten and printed Arabic text, labeling contextual image regions (e.g., storefront signs, license plates, packaging), and tagging image quality for training robust OCR models. Ensured accurate transcription of Arabic content, including dialectal variations, diacritics, and ligature forms. Maintained over 95% QA compliance while processing over 500+ images per batch, helping develop region-aware AI tools for visual recognition in MENA markets.

2024

Education

Work History