All Projects

2D arcade shooter game built with C and Raylib, implementing low-level data structures, game loop mechanics, and persistent game state management.

Result: Led team development focusing on object behavior design, game loop optimization, and user experience management throughout the development cycle.

• Manual implementation of data structures at low level.
• Optimized game loop for responsive gameplay and object behavior.
• Persistent game state and local high-score tracking system.
• Team collaboration and leadership in game development workflow.

Smart photography assistant built with Flutter that combines real-time AI pose estimation (Google ML Kit) and on-device image processing to guide users into better poses and compositions before pressing the shutter.

Result: Delivered a cross-platform camera app with ghost overlay pose guidance, cosine-similarity scoring, live color filters, GPU-accelerated photo editor, and a Supabase-backed cloud pose library accessible after login.

• Real-time ghost overlay skeleton drawn via CustomPainter that color-shifts red → yellow → green as cosine-similarity score surpasses 82%.
• ML pipeline runs Google ML Kit Pose Detection on camera frames in a background isolate, extracting 33 body landmarks per frame with throttling to preserve UI frame rate.
• GPU-accelerated photo editor: adjustments (brightness, contrast, saturation, color temperature) are composed into a single 5×4 color matrix and rendered via Flutter Canvas/PictureRecorder for zero-lag saves.
• Dual storage strategy: Hive (local, offline-first) for custom pose templates and Supabase (PostgreSQL + Storage Bucket) for the cloud public pose library with download-count tracking.
• Clean Architecture layers (Domain / Data / Presentation) with ChangeNotifier controllers, keeping all camera lifecycle, flash, zoom, and capture state out of the UI layer.
• Pose Template Builder: extract pose landmarks from any gallery photo, persist them locally, and browse or share them via the cloud library.

Cross-platform mobile quiz platform built with Flutter that prioritizes network resilience through an offline-first architecture, local Hive storage, background sync, and real-time LAN broadcasting for classroom environments.

Result: Eliminated quiz interruptions in low-connectivity environments by shifting all quiz processing to the client device, enabling seamless end-to-end participation and automatic score sync when the connection is restored.

• True offline-first: entire quiz packages downloaded to encrypted local Hive NoSQL DB, eliminating per-question network calls.
• Background connectivity monitor queues all activity locally and auto-syncs scores to MongoDB when a stable connection is detected.
• LAN Broadcasting mode: Host device acts as a local WebSocket server, enabling real-time multi-participant quizzes over Wi-Fi with zero internet traffic.
• Dual-role system (Host & Participant) with role-specific UIs for quiz creation, management, and live session control.
• Feature-driven architecture with real-time client-side input validation, cryptographic password hashing, and QR-code session joining.

Full-stack integrated sharia cooperative application with complex domain logic handling savings, murabahah-based installments, POS cash system, and inventory management.

Result: Implemented feature-based end-to-end development approach ensuring consistent data flow across financial and inventory modules.

• Complex domain modeling for sharia-compliant financial operations.
• Multi-role access control with centralized authorization system.
• Synchronized data structure across financial and inventory modules.
• Feature-based fullstack development for maintainable codebase.

Cross-platform GUI file manager inspired by Windows Explorer, built in C with raylib/raygui and 4 core ADTs (tree, stack, queue, linked list).

Result: Shipped a desktop file manager for 3 OSes with 10+ keyboard shortcuts, undo/redo, trash recovery, and bulk-operation progress feedback.

• Non-binary tree powers directory hierarchy and sidebar navigation.
• Stack-based undo/redo across create, delete, rename, and paste actions.
• Queue-based clipboard workflow with progress UI for 10+ items.
• Custom window chrome with resizable layout and modern toolbar.
• Cross-platform build flow with bundled raylib and system fallback.

Cross-platform Flutter mobile application implementing digital image processing algorithms and real-time AI Vision, covering image acquisition, color model manipulation, frequency-domain filtering, and object detection with OpenCV.

Result: Delivered an end-to-end mobile lab environment for digital image processing exploration — from camera acquisition and histogram equalization to Fourier-domain filtering and real-time human/pothole detection — with a clean feature-based architecture.

• Onboarding & mock-auth module with rate-limited login lockout for access-control demonstration.
• Full digital image processing pipeline: grayscale, RGB manipulation, histogram, histogram equalization, convolution filters, and sharpening.
• Frequency-domain operations (FFT-based low/high/band-pass filtering) via native OpenCV bindings.
• Real-time AI Vision mode for human entity detection and pothole inspection with bounding-box overlay using CustomPainter.
• Camera Overlay Processor: burn-in composite of live camera frame + AI detection layer saved directly to device storage as a static image file.
• Feature-based folder structure with strict Separation of Concerns (View, Controller, Service, Painter, Model).

Cross-platform desktop productivity application with layered MVC architecture, implementing multiple design patterns for modularity and scalability.

Result: Delivered cross-platform installable application with comprehensive testing and database fallback system ensuring data persistence across operating systems.

• Layered MVC architecture with strict separation of concerns.
• Multiple design patterns for maintainable and scalable codebase.
• PostgreSQL with SQLite fallback stored in cross-platform user directory.
• Comprehensive unit testing and structured input-output validation.
• Focus on UI/UX readability and user workflow optimization.

Interactive 2D pattern-block puzzle built in Godot 4 (C#) with manual 2D transforms, snap-based validation, and custom stage templates.

Result: Shipped a complete gameplay loop with 3 curated 720p challenges, persistent progress + best-time tracking, and a JSON-based Template Builder so users can create and replay patterns.

• Manual 2D transform pipeline plus training-focused mechanics (snapping, hit testing, geometric composition).
• Snap-to-slot validation that locks shapes, triggers 100% completion, and stores persistent progress + best time.
• Full input system: drag-and-drop, Q/E rotation, WASD/arrow micro-move, right-click reset.
• Modular stage/template system (TemplateShapeStage, TemplateLoader) with JSON receipts.
• Template Builder workflow (build -> export JSON -> replay in My Patterns) optimized for desktop Windows/macOS/Linux at 1280x720.