For employers

Hire this AI Trainer

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

LoginCreate account
Invite to Job
M

Mohamed Abdelmohsen Soliman

AI Engineer with years of experience in python and ML

Egypt flagCairo, Egypt
$10.00/hrIntermediateOther

Key Skills

Software

Other

Top Subject Matter

Mental health and wellness
Business document and photo text extraction for AI retrieval
AI Agentic Setups Programming

Top Data Types

TextText
ImageImage
VideoVideo

Top Task Types

Emotion Recognition
Classification
Transcription
Segmentation
Object Detection
Fine Tuning

Freelancer Overview

I have started Data labeling in 2024 when i started my thesis. I was working on image data. It was about facial emotion expressions. I labeled and augmented more than 4000 images manually after capturing it through a 2016 phone camera. I also worked on sentences. It was the same Emotion Classification task. However, this time was through text since I was working on a multimodal model including voice, images, and text. I labeled 7 emotions on more than 800 sentences (the positions where nulls existed). Recently. I worked on state-farm dataset on kaggle for driver distraction. It included 79k unlabeled test images. I labeled 2k of them manually into 10 classes (c0-c9).

IntermediateEnglishArabic

Labeling Experience

LLM Data Labeling and Model Training for RAG Systems (AI Engineer at NeQabty)

TextFine Tuning
I extracted, cleaned, and prepared photo and document-derived text data for use in retrieval-augmented generation (RAG) systems. I deployed and fine-tuned multiple LLM-based models using processed data to support AI retrieval tasks. Responsibilities included crafting representative datasets and ensuring input quality for model performance. • Managed document and photo-derived text annotation and data labeling for LLM-based Azure OpenAI RAG systems. • Created and maintained custom text datasets to enhance model training and retrieval relevance. • Used Azure AI Foundry, LangChain, and internal tools for data processing and labeling workflows. • Focused on business document understanding and automation within RAG pipelines.

I extracted, cleaned, and prepared photo and document-derived text data for use in retrieval-augmented generation (RAG) systems. I deployed and fine-tuned multiple LLM-based models using processed data to support AI retrieval tasks. Responsibilities included crafting representative datasets and ensuring input quality for model performance. • Managed document and photo-derived text annotation and data labeling for LLM-based Azure OpenAI RAG systems. • Created and maintained custom text datasets to enhance model training and retrieval relevance. • Used Azure AI Foundry, LangChain, and internal tools for data processing and labeling workflows. • Focused on business document understanding and automation within RAG pipelines.

2025 - 2025

Nano – S2S Mental Wellbeing Assistant: Emotion Recognition Data Labeling and Model Training (Thesis)

OtherTextEmotion Recognition
I developed a RoBERTa-based emotion classifier for therapy dialogue pairs as part of my undergraduate thesis. I leveraged semantic embeddings and max-voting aggregation to enhance model accuracy and robustness for mental wellbeing assessment. The project involved the use of advanced speech and language models for emotion recognition and natural language understanding. • Labeled and classified emotions in 700k therapy dialogue pairs using custom and pretrained models. • Combined outputs from emotion classifiers via max-voting to improve reliability and reduce noise in annotations. • Integrated Whisper ASR, TTS-1-HD, and GPT-4o for speech/text processing and AI training workflows. • Focused on the mental health and wellness domain, supporting emotion detection tasks for conversational AI.

I developed a RoBERTa-based emotion classifier for therapy dialogue pairs as part of my undergraduate thesis. I leveraged semantic embeddings and max-voting aggregation to enhance model accuracy and robustness for mental wellbeing assessment. The project involved the use of advanced speech and language models for emotion recognition and natural language understanding. • Labeled and classified emotions in 700k therapy dialogue pairs using custom and pretrained models. • Combined outputs from emotion classifiers via max-voting to improve reliability and reduce noise in annotations. • Integrated Whisper ASR, TTS-1-HD, and GPT-4o for speech/text processing and AI training workflows. • Focused on the mental health and wellness domain, supporting emotion detection tasks for conversational AI.

2020 - 2025

Education

T

The American University in Cairo

Bachelor of Science, Computer Engineering

Bachelor of Science
2020 - 2025

Work History

N

Neqabty

AI Engineer

Cairo
2025 - 2025
A

Accor

Information Technology Intern

Sharm-El Sheikh
2024 - 2024