Projects
Selected AI, data science, NLP, accessibility, and analytics projects.
Doctor AI
Explainable Medical AI
A team-built medical AI platform that combines deep-learning models with explainability to support clinical decision-making across multiple diagnostic pathways.
- Contributed to a platform integrating TensorFlow deep-learning models and explainability features across eight diagnostic pathways.
- Focused on practical model integration, evaluation, and responsible AI workflows for healthcare use cases.
- Outcome: Finalist — Jordan Students Pitch Competition 2025.
TensorFlow
Deep Learning
XAI
Medical AI
Noor
Assistive AI
A real-time accessibility application designed to support blind and deaf users through computer vision, OCR, sign-language conversion, and caregiver support.
- Developed the concept and prototype around scene description, OCR, sign-language interaction, and a caregiver dashboard.
- Combined accessibility-focused AI features into a single practical user workflow.
- Outcome: 2nd Place — Mafraq Governorate AI Hackathon 2025.
Computer Vision
OCR
Accessibility
AI
Mowthooq
NLP + Verification
An AI product concept for content creators focused on news verification, spelling support, text-to-speech, and more responsible content workflows.
- Designed the product workflow around content verification and AI-assisted authoring.
- Explored NLP, LLM-assisted verification, language support, and accessibility through text-to-speech.
- Presented during the Tech3arabi AI Hackathon.
NLP
LLMs
Verification
TTS
U.S. Hospital Performance & Health Tourism Dashboard
Power BI
An interactive analytics dashboard combining public U.S. healthcare and tourism data to explore hospital performance and health-tourism trends.
- Used public CMS and International Trade Administration datasets.
- Performed data cleaning, modeling, Power Query transformations, and DAX measure development.
- Analyzed hospital quality, safety, patient experience, and tourism spending through interactive visuals.
Power BI
Power Query
DAX
Data Modeling
English → Arabic Machine Translation
NLP
Fine-tuned a pretrained Transformer on parallel English/Arabic text and evaluated translation quality with SacreBLEU.
- Dataset: OPUS-100 English/Arabic parallel corpus.
- Tooling: Hugging Face Transformers, Datasets, Evaluate, and Gradio.
- Training: Seq2SeqTrainer with a pretrained sequence-to-sequence model.
Transformers
Hugging Face
SacreBLEU
Gradio
GrowWise
EdTech
App concept that personalizes students’ technology learning journeys by identifying skill gaps and recommending courses and projects.
Product Design
Recommendation
Roadmapping