Hi, I'm Moubarak Ali-Kparah

Core Architecture & Privacy-First Overlay: Engineered a high-performance Privacy-First Overlay using Electron, implementing advanced window management (transparency, click-through forwarding, and always-on-top states) to provide non-intrusive AI assistance. I recently improved this by implementing zero-latency rendering using ready-to-show callbacks to solve rendering bugs, and added a distraction-free interaction mode that seamlessly integrates with OS workflows while maintaining data security.

Resilient Multi-Provider LLM Orchestration: Originally architected as a 7-layer "Resilient Multi-Provider LLM Controller" routing system, I have since expanded this to a massive 18-model logic engine. This system now intelligently cascades across tiers—ranging from supreme reasoning models like Gemini 3.0 Pro and Claude 3.7 to efficiency models like DeepSeek—ensuring zero downtime. To stabilize this, I engineered a history sanitization slicer to prevent API rejections and a vision-based "Solve" tool that automatically routes traffic to the correct Vision or Text brain based on the payload.

Real-Time Multi-Modal Data Streams: Developed a hybrid speech-to-text system using a local Whisper server with a custom restart loop to process audio in continuous chunks, effectively bypassing API rate limits. To further refine this, I implemented a ticket-based sorting system that resolves race conditions where cloud responses might arrive out of order, ensuring perfect chronological transcriptions. I also added a 1000ms cloud fallback logic to trigger whenever the local server queue is over-taxed.

System-Level Features & Optimization: Beyond global keyboard shortcuts (Ctrl+H) and native Tray APIs, I have optimized the application’s footprint through RAM-based smart caching, which reads and stores student profiles once on startup for instant retrieval. I also refactored the underlying architecture to use type-only imports, successfully resolving circular dependency crashes that previously caused startup hangs, and integrated live terminal telemetry to monitor IPC and LLM data flow in real-time.

Built a reactive web application leveraging Appwrite for secure authentication and real-time database management to track and visualize trending content from external REST APIs.

Rich Detail Display: Designed a custom modal architecture to display detailed movie information (overview, rating, vote count) on click, preserving user context.

Intelligent Streaming Routing: Engineered a smart traffic routing system that dynamically detects the user's device:
— PC/Web: Links directly to the web streaming site, Nkiri.
— Mobile (Android/iOS): Links to Nkiri, but also offers an exclusive option for the native app, Anilab, through optimized mobile deep-linking (Android Intents).

Designed a fully responsive interface using Tailwind CSS, creating a custom modal architecture that displays rich movie details and preserves user context across mobile and desktop devices.

Developed a high-performance portfolio site using modern CSS Grid and Flexbox, achieving a 100/100 Lighthouse performance score.
Automated deployment pipelines via GitHub Actions with custom domain configuration and SEO optimization.
Implemented a custom Light/Dark Mode toggle with a modern "Silver/Concrete" light theme and "Neon/Dark" mode.
Features Interactive Animations like custom shine effects, hover scaling, and continuous loop animations on skill bars.
Designed for Accessibility, including optimized focus states, semantic HTML, and high-contrast colors.