Mohamed Abdelgawad

Key Skills

Software

AWS SageMaker

CVAT

Encord

Labelbox

Label Studio

Scale AI

SuperAnnotate

Top Subject Matter

No subject matter listed

Top Data Types

3D Sensor

Image

Text

Top Task Types

Bounding Box

Classification

Entity Ner Classification

Prompt Response Writing SFT

Segmentation

Freelancer Overview

I'm a bilingual AI data specialist (Arabic & English) with 3 years in labeling, RLHF evaluation, and analytics. I maintained a 95%+ QA scores across more than 30K text and image judgments, earning early access to red-team and GPT-4 ranking pilots. Recent projects: LLM prompt-tuning for a deal-closure model that boosted F1 from 0.71 to 0.83 in two weeks. Building a Power BI pipeline that cuts loan portfolio reporting by 50%. Leading a Sentiment Analysis capstone included manually labeling 10k chat snippets and crafting sentiment/flight-risk features.

Labeling Experience

Code Snippet Review & Categorization for LLM Training

Scale AITextClassificationEvaluation Rating

Evaluated and labeled thousands of code snippets and code generation prompts for tuning a LLM coding capabilities. Tasks involved classifying Python/JavaScript snippets by function and tagging prompt quality.

Evaluated and labeled thousands of code snippets and code generation prompts for tuning a LLM coding capabilities. Tasks involved classifying Python/JavaScript snippets by function and tagging prompt quality.

2023

Bilingual Audio Transcription & Sentiment Tagging

Label StudioAudioClassificationTranslation Localization

Transcribed 100 h of Arabic to English support calls and applied sentiment labels. Dataset trimmed ASR post-edit time by 35%.

Transcribed 100 h of Arabic to English support calls and applied sentiment labels. Dataset trimmed ASR post-edit time by 35%.

2024 - 2024

Employee Chat Sentiment & Flight-Risk

LabelboxTextEntity Ner ClassificationClassification

Labeled 10k messages with sentiment (pos/neg/neu) and flight-risk indicators; tagged entities for department linkage. The resulting set powered a regression model that flagged at-risk teams.

Labeled 10k messages with sentiment (pos/neg/neu) and flight-risk indicators; tagged entities for department linkage. The resulting set powered a regression model that flagged at-risk teams.

2024 - 2024

Retail Image Annotation for Product Recognition

CVATImageBounding BoxClassification

Annotated 8k high-resolution shelf photos, boxing individual SKUs and linking eavh box to 120-class product. Achieved 97% QA pass rate.

Annotated 8k high-resolution shelf photos, boxing individual SKUs and linking eavh box to 120-class product. Achieved 97% QA pass rate.

2024 - 2024

Education

Work History