AI Engineer @ Mawhub · Software Engineering @ Umm Al-Qura University '27 · Competitor @ WorldSkills Shanghai 2026
I build software and applied AI systems with an emphasis on reliable implementation, evaluation, testing, and deployment. My work has ranged from privacy-preserving NLP on Raspberry Pi hardware to full-stack software engineering and international software-testing competition.
Currently, I work as an AI Engineer at Mawhub, where I take engineering work from requirements and technical decisions through implementation, testing, documentation, and integration.
- WorldSkills Shanghai 2026 — Software Testing: represented Saudi Arabia after approximately four months of focused preparation across web, mobile, API, performance, and white-box testing.
- KAUST Academy AI: selected among the top 100 students from 14,000+ applicants and completed the AI Specialisation and eight-week summer internship.
- OnKith: developed the privacy-model track from early sequence-tagging baselines to an INT8 DeBERTa deployment evaluated on IID, OOD, hard-negative, and Raspberry Pi benchmarks.
- Software engineering: coordinated a six-person graduation project and worked across requirements, backend/frontend implementation, testing, documentation, and repository governance.
OnKith · privacy-preserving edge AI
OnKith is a privacy-first local voice-processing system designed to remove personally identifying information before sensitive text leaves the device. I focused on the privacy-model, data, evaluation, and deployment track.
I helped evolve the privacy model from BiLSTM → TinyBERT → DeBERTa-v3-xsmall, built stronger leakage-aware data and evaluation protocols, and productionised the final model through ONNX and INT8 quantisation.
- 284,619 English rows and 2,088,335 labelled spans across 31 entity types
- Frozen OOD typed F1 improved from 0.4649 → 0.6244
- Private-character recall improved from 0.3209 → 0.8390
- Hard-negative false-positive rate reduced from 0.6528 → 0.2361
- Raspberry Pi 5 text benchmark: 0.9448 typed F1 at 62.4 ms median masking latency
PyTorch Transformers DeBERTa ONNX Runtime INT8 Token Classification Raspberry Pi
ESAS · graduation project
Experience Saudi As a Saudi is a backend-first tourism marketplace prototype for locally curated Saudi experiences. I coordinated the six-person team, owned the repository and major project documentation, reviewed contributions, and contributed to implementation across the system.
The prototype uses Java 21, Spring Boot, PostgreSQL, Flyway, Flutter, JWT-based security, REST APIs, and Docker, with role-aware flows for travelers, providers, and administrators. Implemented areas include catalog browsing, authentication, cart and simulated checkout, booking flows, wishlist, provider onboarding/submission, and admin approval/moderation.
Java Spring Boot PostgreSQL Flutter REST APIs Docker Software Engineering
kaust-cell-instance-segmentation · computer vision
Placed 3rd of 24 teams in a KAUST Academy cell-instance segmentation challenge. The final approach used a ConvNeXt-Tiny U-Net, continuous per-instance distance prediction, object-centric sampling, and marker-controlled watershed.
Final private leaderboard score: 0.5472 · grouped-validation F1: 0.8144
PyTorch ConvNeXt U-Net OpenCV Watershed
Kaggle_inpainting_comp · image inpainting
Built an MI-GAN reconstruction pipeline with test-matched dynamic masks, YuNet face filtering, fixed evaluation masks, and gated checkpoint selection.
Reconstructed all 8,000 test images and finished 3rd with a private-leaderboard FID of 12.02.
PyTorch MI-GAN OpenCV YuNet Clean-FID
flappy_bird_challenge · reinforcement learning
Built a CPU-trained Dueling Double DQN with prioritised experience replay and stability-focused checkpoint selection. Diagnosed a major training collapse from evaluation history and rebuilt the target-update strategy using Polyak averaging.
Best episode: 5,120 pipes · unseen five-seed holdout mean: 400
PyTorch Gymnasium NumPy
Competition preparation and hands-on work with:
Selenium Appium Postman Newman JMeter pytest
Across:
Web testing · Mobile testing · API testing · Performance testing · Automated testing
Languages: Python · Java · TypeScript · HTML/CSS
AI / ML: PyTorch · Hugging Face Transformers · ONNX Runtime · TensorFlow Lite · OpenCV · NumPy
Software: React · Vite · Tailwind CSS · REST APIs · Docker · Git · GitHub
Testing: Selenium · Appium · Postman · Newman · JMeter · pytest