Ahmed Salah

Key Skills

Software

Appen

Data Annotation Tech

Remotasks

Roboflow

SuperAnnotate

Scale AI

Top Subject Matter

No subject matter listed

Top Data Types

Audio

Document

Text

Top Task Types

Audio Recording

Data Collection

Prompt Response Writing SFT

Text Generation

Text Summarization

Freelancer Overview

I have hands-on experience in data labeling and AI training data, working across multiple projects that involved audio transcription, image classification, and natural language data annotation. My background includes contributing to high-quality datasets used for training large-scale language models, ensuring accuracy, consistency, and adherence to project-specific guidelines. I have worked with tools like [mention tools if any, e.g., Labelbox, SuperAnnotate, or custom platforms], and I am familiar with annotation taxonomies, quality assurance practices, and following detailed labeling instructions. What sets me apart is my strong attention to detail, ability to quickly adapt to new guidelines, and my experience in both independent and collaborative labeling tasks. I am comfortable working with various data types and always strive to meet quality and productivity targets. My communication skills and reliability have helped me succeed in remote, deadline-driven environments, and I’m always eager to learn and contribute to impactful AI development. For example, I contributed to the Xylophone Grassland Project, focusing on voice data labeling in Egyptian Arabic, where I was responsible for generating and validating natural-sounding prompts across different categories.

Labeling Experience

Dogwood – Arabic Prompt Generation & Response Writing

Scale AITextClassificationText Generation

In the Dogwood project, I contributed to the development and fine-tuning of a large language model (LLM) through supervised fine-tuning (SFT) in Arabic. My tasks involved crafting high-quality prompts and responses across various domains (e.g., general knowledge, daily life, opinion-based questions). I ensured that the prompts were diverse, realistic, and linguistically sound, while the responses maintained coherence, helpfulness, and alignment with AI safety and ethical guidelines. I also participated in the review process, providing feedback and quality ratings to other contributors' work. The project required a deep understanding of context, cultural sensitivity, and writing clarity in Modern Standard Arabic and Egyptian dialect.

In the Dogwood project, I contributed to the development and fine-tuning of a large language model (LLM) through supervised fine-tuning (SFT) in Arabic. My tasks involved crafting high-quality prompts and responses across various domains (e.g., general knowledge, daily life, opinion-based questions). I ensured that the prompts were diverse, realistic, and linguistically sound, while the responses maintained coherence, helpfulness, and alignment with AI safety and ethical guidelines. I also participated in the review process, providing feedback and quality ratings to other contributors' work. The project required a deep understanding of context, cultural sensitivity, and writing clarity in Modern Standard Arabic and Egyptian dialect.

2025

Xylophone Grassland – Arabic Voice Data Collection & Labeling

Scale AIAudioText GenerationEmotion Recognition

This project involved generating, recording, and annotating Arabic voice data, specifically in Egyptian dialect, to train and fine-tune large language models (LLMs). Tasks included writing natural conversation prompts across categories (casual, knowledge-based, and roleplay), recording voice samples, transcribing spoken text, labeling conversation types and emotions, and reviewing peer submissions for quality. I worked on hundreds of samples while ensuring adherence to strict linguistic, timing, and formatting guidelines. The project demanded high attention to detail, consistency, and creativity in designing realistic dialogue scenarios. All recordings were evaluated based on clarity, tone, and prompt alignment.

This project involved generating, recording, and annotating Arabic voice data, specifically in Egyptian dialect, to train and fine-tune large language models (LLMs). Tasks included writing natural conversation prompts across categories (casual, knowledge-based, and roleplay), recording voice samples, transcribing spoken text, labeling conversation types and emotions, and reviewing peer submissions for quality. I worked on hundreds of samples while ensuring adherence to strict linguistic, timing, and formatting guidelines. The project demanded high attention to detail, consistency, and creativity in designing realistic dialogue scenarios. All recordings were evaluated based on clarity, tone, and prompt alignment.

2025 - 2025

Education

Work History