Computer Vision · Multimodal AI · LLM Systems & Evaluation
I build AI systems for visual understanding, retrieval, and evidence people can inspect. My work spans model development, evaluation, and integration into applications.
LinkedIn · Email · Selected work · Research
Madinah, Saudi Arabia · B.Sc. Artificial Intelligence, University of Prince Mugrin · Expected December 2026
Multimodal hiring integrity with timestamped observations and evidence for human review. My contribution Best Prototype, 1st Place |
Compromise assessment across Windows Event Logs and memory evidence, with findings reviewed by an analyst. My contribution 94.1% micro F1 on Event Logs |
Hand tracking, real time physics, and WebGL. An individual project exploring spatial interaction, with documented experiments, interface imagery, and an open account of the engineering work still needed.
MediaPipe · Three.js · WebGL · Explore the experiments and prototype
 |
Quran & Hadith Retrieval Arabic and English search across 40,098 records. My work: Arabic normalization, hybrid retrieval, and the evaluation harness. Corpus size describes scale; retrieval quality evaluation is pending. |
 |
Smart Cashier Grocery detection connected to a reviewable checkout. My work: dataset preparation, detector training, and Streamlit integration. Reported mAP@0.5: 0.9696 on the original validation split. |
 |
Skin Lesion Study Classification under extreme class imbalance in the ISIC 2024 setting. My work: CNN and image preprocessing. Academic team study; no published performance benchmark. |
The covers are concept illustrations. Each project page identifies its public material, contribution scope, and evaluation limits. Source code remains private.
| Area | Selected tools and methods |
|---|---|
| Vision and multimodal | PyTorch, OpenCV, InsightFace, YOLOv8, MediaPipe, WavLM |
| LLM systems and retrieval | Local model serving, Sentence Transformers, FAISS, structured outputs |
| Applications and data | FastAPI, PostgreSQL, SQLAlchemy, AWS S3 |
| Evaluation | Benchmark design, ground truth, threshold analysis, retrieval and detection metrics |
AUT4M · Expert System Supporting Autism Screening in Children
Lead author of a working paper on a GARS based screening support methodology, developed collaboratively and trialled under psychologist supervision. Designed for screening support, not standalone diagnosis.
Earlier work and development
Pix-alchemist (2024). Image enhancement for 600 degraded grayscale images of Madinah date varieties, refined through more than 270 controlled experiments.
Project Bayan (2023). Team robotics project involving motion sequencing, voice interaction, assembly, and iterative testing in a five day programme. The prototype was demonstrated to HRH Prince Muqrin bin Abdulaziz and HRH Prince Faisal bin Salman, with national television coverage.
| Year | Development |
|---|---|
| 2023 | Humanoid Robot Programming, Smart Methods |
| 2022 | Deep Learning Specialization, DeepLearning.AI |
| 2022 | AI Summer Champions Program, SDAIA |
Open to AI engineering opportunities and research collaboration.
Arabic native · English professional