Derbent is an edge-native Single Sign-On (SSO) and authentication engine built specifically for the Cloudflare ecosystem. I built this to solve the complexity and latency issues associated with traditional authentication setups, aiming to create a centralized, self-hosted service that securely manages user sessions across multiple subdomains.
What I Built
I engineered a highly secure, stateful authentication backend from the ground up using Cloudflare Workers, Cloudflare D1 (SQLite), and KV storage. Instead of relying on stateless JWTs—which are notoriously difficult to revoke—I designed a stateful session architecture. This ensures immediate session invalidation and superior security, all while running with zero cold starts at the edge, globally closer to the user.
Key Features & Technical Achievements:
- Edge-Native Performance: Deployed entirely on Cloudflare Workers, ensuring millisecond latency and high availability without managing traditional servers.
- Stateful SSO Architecture: Engineered a cross-subdomain Single Sign-On system that securely authenticates users across a suite of microservices seamlessly.
- Advanced Security Measures: Built-in session hijack protection, IP and User-Agent tracking, and detailed audit logging to maintain high security standards.
- Service Binding Integration: Utilized Cloudflare’s Service Bindings to allow internal apps to securely verify sessions with zero network overhead.
- Seamless OAuth: Integrated GitHub OAuth for frictionless user onboarding and login.
The Impact
This project highlights my deep understanding of backend infrastructure, modern edge computing, and application security. It demonstrates my ability to design scalable, centralized microservices and tackle complex architectural challenges—like secure cross-domain authentication—without relying on heavy, off-the-shelf third-party providers.
0
53
Nâmedâr is an offline-first Optical Mark Recognition (OMR) platform designed for educators. I built it to solve a major problem in EdTech: the privacy risks and latency of uploading sensitive student exam data to cloud servers for grading.
What I Built
I engineered a full-stack solution that allows teachers to create, scan, and instantly grade paper exams entirely within their web browser. To achieve this without relying on backend servers, I developed a high-performance computer vision engine using Rust and WebAssembly (WASM). This handles complex image perspective correction and grading locally on the user's device, guaranteeing a "zero-upload" privacy model where student data never leaves the computer.
Key Features & Technical Achievements:
- Advanced In-Browser Processing: Leveraged Rust and WASM to bring heavy optical scanning computations directly to the frontend, resulting in lightning-fast, offline grading.
- Modern Frontend Architecture: Built a highly responsive UI using React and TanStack Start.
- Intuitive Form Builder: Implemented a drag-and-drop interface allowing educators to design custom bubble sheets easily.
- Data Portability: Engineered multi-format data exports (CSV, JSON, XML) so teachers can seamlessly sync grades with their school's existing systems.
The Impact
This project demonstrates my ability to take a complex, backend-heavy concept (computer vision/grading) and successfully migrate it to the frontend. It showcases my expertise in combining modern web frameworks (React) with systems-level languages (Rust) to build fast, secure, and privacy-first web applications.
0
69
🍕 Pizza Lab — A Visual Pizza Design Studio
🔗 Live Project: https://sweat-bunch-55404838.figma.site/
(https://sweat-bunch-55404838.figma.site/)🔗 Figma Community File: https://www.figma.com/community/file/1649547758127634830
The Problem
Digital food ordering is often a static, uninspiring experience. You click checkboxes on a text list and just hope the food looks good when it arrives. I built Pizza Lab to solve this by turning the ordering process into an immersive, highly visual design experience.
The Workflow: Built with Figma Make & Weave
To bring this highly interactive UI to life, I utilized a full Figma AI workflow:
- Asset Generation: I used Figma Weave to generate the highly realistic pizza layers (crusts, sauces, cheeses, toppings) and the chef's expressions. I then processed these and hosted them via Cloudflare R2 for perfect top-down stacking.
- Code Generation: I designed the core interface components—including the ingredient carousel, the recipe card, and the dynamic feedback module—in Figma. Using Figma Make, I prompted the designs directly into fully functional, production-ready UI code. This allowed me to focus heavily on the complex app logic (state management, CDN integration, and Framer Motion animations) rather than writing boilerplate code.
Key Features:
- Live Layer System: Four stacked layers (Crust → Sauce → Cheese → Toppings) with a live top-down orthographic preview.
- Ultra-Realistic Details: Features per-ingredient random rotation for visual variety, a subtle heat shimmer animation on the pizza, and realistic drop shadows.
- Keyboard First: Fully navigable via keyboard (↑/↓ to switch layers, ←/→ to navigate the smooth-spring ingredient carousel, Space to select).
- Dynamic UI: 4 table textures (Dark Marble, Walnut, Oak, Light Marble) where the UI text automatically adapts its color to remain readable on both light and dark surfaces.
- Chef Feedback: An animated chef character reacts to your ingredient pairings (e.g., looks disappointed at pineapple), featuring quotes from famous Neapolitan pizza masters.
- Export & Share: Randomize gourmet combinations, save the exact recipe as a JSON file, or encode the exact pizza state into a shareable URL!
1
195