Portfolios
Ayar+ | عياربلاس
→
Overview
Ayar+ is a full-featured .NET MAUI client for jewelry and precious-metals businesses across the Middle East and South Asia. It connects to a Laravel REST API and gives shop owners and accountants one app to manage invoices, bills, receipts, contacts, inventory, warehouses, currencies, financial reports, and user permissions - scoped per workspace (store/branch). The app ships in 6 UI languages with full RTL layout for Arabic and Persian, and integrates Civil ID card readers, barcode/QR login, realtime Ably events, PDF/Excel export, and OneSignal push notifications.
Objective
Deliver a reliable, multilingual, permission-aware client that lets jewelry traders create complex gold-weight documents with live gold rates, supports multi-workspace tenancy with correct API scoping, works on phones/tablets/desktop (TabBar on mobile, Flyout on Windows/Mac), and integrates with hardware ecosystems including Kuwait PACI Civil ID readers and barcode scanners.
Challenges and Solutions
Challenge: Multi-tenant workspace scoping evolved from company_id to workspace UUIDs with three identifier types coexisting. Solution: Centralized preference keys, AuthenticationHandler auto-attach of X-Workspace header, legacy migration at startup, and documented API drift for backend coordination. Challenge: Complex jewelry document calculations (gold weight, karat conversion, stone weight, wage percent, live rates). Solution: Dedicated ViewModels per document flow, AyarConversionService, CachedGoldRateService with stale-on-error fallback, and DocumentDraftSessionStore for in-progress drafts. Challenge: Cross-platform PDF preview/export with RTL Arabic text and custom fonts. Solution: PdfSharpCore + SkiaSharp font rendering with platform-specific rasterizers and custom Android WebView PDF facades. Challenge: Realtime Civil ID card reader integration between Windows desktop and mobile invoice flows. Solution: Ably channel subscription with debounced CivilIdScannerService and typed payload mapping for contact/invoice prefill. Challenge: Permission-aware navigation at scale (40+ routes, role hierarchy). Solution: PermissionGuardService, PermissionKeys synced with backend, and AppShellViewModel visibility properties per tab/route.
Role and Takeaways
Led development of Ayar+ as the primary mobile client - owned architecture from the HTTP pipeline (named HttpClient, Bearer auth, tenant headers) through Shell navigation to PDF/Excel export. Built 20+ typed API service integrations with 80+ DTOs, 45+ ViewModels, and 42+ pages. Shipped Android releases via GitHub Actions (AAB/APK with Sentry monitoring), established localization/RTL standards across the UI, and documented API/client contract gaps for backend coordination.
App Gallery
Gate Access Manager
→
Overview
A robust vehicle access system designed for sequential gate passage. Integrated with logging, rule-based filters, and detailed record handling.
Objective
To streamline controlled access with high configurability, device support, and maintainable data operations.
Challenges and Solutions
Challenge: Handling complex, dynamic database structures and custom request rules. Solution: Designed normalized relational models with robust filtering and validation logic.
Role and Takeaways
Independently implemented the software layer and database structure. Learned advanced SQL design and high-reliability gate logic handling.
SpeedCam Dispatcher (VSD Package)
→
Overview
Composed of watchdog, transmitter, updater, and ViolationSystemDetector (VSD), each system processed two camera feeds using darknet-based YOLOv4 models under CPU-only constraints.
Objective
To capture, detect, and transmit traffic violations using weak AI models with limited hardware resources.
Challenges and Solutions
Challenge: Real-time detection with weak CPUs, integrating YOLOv4 in C++, not Python. Solution: Customized and embedded darknet model in C++, used OpenVINO for inference optimization, achieving detection in 10FPS on 7th-gen Core i5.
Role and Takeaways
Built and optimized the detector and transmission modules; designed custom detection and preprocessing algorithms; gained deep insights into AI deployment limits.
Deviman Portfolio Website
→
Overview
Features a responsive, modern design with dynamic content rendering, interactive UI components, and seamless data management. Implemented secure user authentication and authorization, and optimized for performance and scalability.
Objective
To showcase my expertise in full-stack web development, including backend logic, database interaction, and responsive frontend design, while providing a platform to present my projects and skills.
Challenges and Solutions
Challenge: Implementing robust data persistence and complex query operations with Entity Framework Core; ensuring cross-browser compatibility and optimal performance for dynamic content. Solution: Designed a normalized SQL Server database schema, utilized EF Core for efficient data access, and employed asynchronous programming for improved responsiveness. Addressed compatibility issues through extensive testing and implemented client-side caching strategies.
Role and Takeaways
Led the end-to-end development, from database design and API implementation to frontend UI/UX. Gained significant experience in architecting scalable ASP.NET Core applications, optimizing database interactions, and delivering a polished, user-centric web experience.
Mobile Traffic Enforcer
→
Overview
Captured and transmitted double-parking and illegal parking violations through SIM-card-connected networks using offline caching.
Objective
To bring mobile enforcement capability to field police using a familiar and offline-capable interface.
Challenges and Solutions
Challenge: Offline data caching with slow wireless connections. Solution: Used SQLite and JSON export with later upload mechanism to match organizational backend.
Role and Takeaways
Developed the app interface, data logic, and transmission flow. Extended VSD logic to mobile WinForms environment.
Vehicle & OCR Module
→
Overview
VehicleDetectionEngine handled full-frame detection using OpenVINO-accelerated YOLO models. SmartOCRModule focused on cropped regions for plate OCR using Tenserflow and OpenCV.
Objective
To provide lightweight, fast detection and OCR functionality as embedded modules in traffic systems.
Challenges and Solutions
Challenge: Achieving usable performance without GPU, working on cropped B&W and color blobs. Solution: Optimized both modules for CPU-based inference and introduced dynamic region-of-interest filtering.
Role and Takeaways
Developed, optimized, and maintained both modules. Reused across all major traffic detection systems as core engines.
Vehicle Classifier Engine
→
Overview
Started from 1,000 labeled samples, used a self-augmenting pipeline with real-world data captured by VSD services to scale to 59,000 samples.
Objective
To classify vehicle types in real-time to aid traffic violation analytics and reports.
Challenges and Solutions
Challenge: Lack of training data and manual labeling limitations. Solution: Designed a semi-supervised feedback loop to auto-capture and label from production data.
Role and Takeaways
Built the dataset, training pipeline, and model integration. Demonstrated practical ML deployment with limited labeled data.
Data Transmitter
→
Overview
Responsible for storing and sending encrypted JSON payloads of violation data, retrying failed entries and categorizing logs.
Objective
To ensure accurate and orderly transfer of traffic data under unstable or limited networks.
Challenges and Solutions
Challenge: Managing duplicates, retries, and failed transmissions without database corruption. Solution: Used SQLite for local queueing and logs, and managed directory-based file separation.
Role and Takeaways
Upgraded and Refined the data transmission logic. Ensured reliability and compatibility with external ASP.NET APIs.
Camera Calibration + Updater
→
Overview
Helped generate and distribute calibration files, aligning detection zones across camera channels and updating software remotely.
Objective
To synchronize and maintain performance for camera-linked detection systems across city-wide deployments.
Challenges and Solutions
Challenge: Handling dual-channel image differences and device updates remotely. Solution: Designed intuitive UI for calibration, and wrote lightweight update services installed on all VSD-linked machines.
Role and Takeaways
Created and deployed both service and UI tool. Learned about camera mapping logistics and centralized maintenance strategies.