I build reliable AI systems that make evidence, uncertainty, and human review visible.
I am a software engineering student at MedTech working at the intersection of machine learning, multimodal systems, interpretability, and human-supervised automation. My work focuses on a practical question: when an AI system gives an answer, can people inspect the evidence, understand its uncertainty, and decide when human review is needed?
- Building evidence-grounded LLM workflows for property management at Joya
- Researching how developer responsiveness shapes user trust in AI applications
- Developing multimodal document-verification systems using text, layout, signatures, and stamps
- Interested in faithful explanations, robustness, model evaluation, and AI systems that fail safely
| Area | Questions I care about |
|---|---|
| Interpretable machine learning | Are the explanations shown to users faithful to the model's real decision process? |
| Multimodal reasoning | How do models behave when textual and visual evidence is missing, unclear, or conflicting? |
| Reliable AI systems | How can evidence checks, uncertainty signals, and human approval gates reduce harmful errors? |
| Evaluation | How can reproducible datasets and failure-case analysis expose what aggregate metrics hide? |
|
A multimodal official-document verification system for diplomas and transcripts. It combines token-layout representations with signature and stamp detection, then maps predictions into inspectable evidence and risk indicators. Focus: PyTorch · LayoutLMv3 · computer vision · explainability |
A fully local assistant that answers questions from PDF documents and interior-design images through retrieval and vision-language pipelines, exposed through a FastAPI backend. Focus: LangChain · FAISS · Hugging Face · FastAPI |
|
A role-based career opportunities platform connecting students, universities, and companies across internships, research positions, and graduate roles. Focus: React · Node.js · Express · MongoDB · JWT |
A privacy-conscious driver-risk prototype combining phone sensors, camera signals, and OBD-II vehicle data to detect distraction, fatigue, and risky driving patterns. Focus: sensor fusion · on-device AI · risk scoring · responsible mobility |
Machine learning and vision
Backend, data, and automation
Application development
15 public repositories · Top languages: JavaScript, CSS, Swift, Python, HTML
| Recently active repository | What it explores |
|---|---|
| UniMatch | MERN platform connecting students, universities, and companies for internships, research, and graduate jobs |
| AUTHENTIQA | ISS Class Project |
| mashroom-genai-rag | AI assistant that answers questions about PDFs and interior design images |
| JE-Hackathon | JE Hackathon |
| CabinKit | CabinKit |
Generated automatically from public GitHub data.
- SMU Foundation Excellence Scholar
- Thomas Jefferson Scholar, U.S. Department of State
- First place, Entrepreneurs of the Future, for the Abyar smart-water project
- Captain of the 2025 JCI World Congress English Debate Championship runner-up team, the first global debate final in JCI Tunisia's history
- Languages: Arabic, English, and French