My Projects
Architecting the future of Agentic AI, Autonomous Systems, and Full-Stack Engineering.
Technical Stack
Project Goal
To create a digital presence that matches the elite quality of Little Pockets' work, providing a seamless booking and service discovery experience for luxury car owners and commercial fleet managers.
Key Features & Deliverables
Premium Aesthetics & Branding
Designed a high-end, visual-first interface that reflects the luxury and precision of elite vehicle customization.
Service Showcasing
Developed interactive sections for luxury wraps, paint protection film (PPF), ceramic coatings, and commercial graphics.
Performance & SEO
Optimized for lightning-fast load times and localized SEO to dominate the Michigan vehicle customization market.
Mobile-First Experience
Ensured 100% responsiveness for customers browsing exotic car transformations on the go.
Technical Stack
Project Vision
Instagram Pro Tools is a high-performance Chrome Extension (Manifest V3) designed to strip away intrusive UI elements and provide a "Theater Mode" for an optimized Instagram desktop experience. It tackles the challenges of modifying a complex, obfuscated React-based DOM to deliver a cleaner, more immersive user interface.
Engineering Challenges & Solutions
Theater Mode (UI/UX Optimization)
Engineered targeted CSS selectors to identify and modify Instagram's complex React-based modal structures, expanding media containers to 100% width while preserving site integrity.
Persistent State Management
Implemented a robust feature toggle system using chrome.storage.local and chrome.scripting, ensuring user preferences (Theater Mode, Play Button Hider) persist across browser sessions.
The 'Shotgun' CSS Strategy
Developed a resilient CSS injection method to neutralize obfuscated class names and persistent video overlays (Reels, Feed Posts), solving the 'DOM-Whack-A-Mole' challenge.
Sleek Popup Controller
Designed a modern, dark-mode control panel using Vanilla CSS and Asynchronous JavaScript, providing real-time visual feedback via dynamic icon badges and color-coded status indicators.
Tech Stack
Project Vision
The era of manual searching is over. OmniCart is a visual intelligence engine that bridges the physical and digital worlds, transforming your phone into an autonomous economic agent. We are moving from "searching" for products to "commanding" outcomes.
Key Features & Roadmap
Phase 1: The Visual Scout (V1)
"The Shazam for Shopping."
- Visual Search: Snap a photo of any item, and OmniCart's AI identifies it.
- AI-Powered Aggregation: The engine scrapes the global web for the best matches, prices, and alternatives in real-time.
- Frictionless Experience: Built on a secure, cross-platform architecture for a seamless experience on both iOS and Android.
Phase 2: The Agentic Awakening (V2)
Your AI agent takes action.
- Autonomous Execution: Set your price and command your agent to execute the purchase when the conditions are met.
- Predictive Logistics: The agent anticipates your needs, queuing up actions for your approval.
- Negotiation Bots: In the future, AI agents will negotiate prices with sellers on your behalf.
Technical Stack
Technical Achievement
AuraDesk bridges the gap between traditional telephony and agentic AI. The V2 refactor transitioned the system from a niche auto-shop bot into a 100% industry-agnostic engine, featuring carrier-level recording, secure audio proxying, and global resource pooling for sub-5s response latency.
Engineering Challenges & Solutions
Multi-Tenant SaaS Engine
Re-engineered the Python backend into an industry-agnostic engine. Developed dynamic instruction injection that merges business-specific personalities with universal safety guardrails.
LLM Precision Migration
Architected the shift to GPT-4o-mini, achieving 99% reduction in tool hallucinations and significantly improving function-calling precision for complex scheduling.
Zero-Lag Tool Logic
Implemented asyncio.wait_for "Hard Walls" across Calendar integrations to prevent dead-air silences and ensure consistent <5s response times.
The "Bouncer" Middleware
Custom IP-reputation system monitoring sensitive paths (/.env, /backup) with tiered automated banning and proxy-aware whitelist detection.
Secure Audio Proxying
Developed Next.js 15 API routes to proxy private Twilio media, allowing dashboard playback without public exposure of carrier-level credentials.
Decisive Disconnect
Engineered single-turn disconnect logic that triggers the hangup tool in the same response as final confirmation, eliminating post-booking silence.
Key Takeaways
Optimized voice fluidity with "Hard Wall" timeouts for external APIs, protected system integrity via the "Bouncer" middleware, and delivered a premium "Supernova" UX with real-time ROI tracking and analytics integration.
Technical Stack
Project Goal
To resolve critical mobile layout bugs and significantly improve site performance and Core Web Vitals for an e-commerce platform with a large asset library of over 17,000 images.
Key Accomplishments & Problem-Solving
Responsive Design & Mobile UI Repair
Performed deep CSS debugging to resolve media query conflicts, ensuring 100% visibility of the logo and mobile menu on all mobile devices and browsers.
High-Scale Next-Gen Image Migration
Executed a background optimization strategy to convert over 17,000 images to WebP format, with critical asset exclusions to maintain brand integrity.
Overcoming Advanced Server Security Blocks
Engineered a bypass using client-side JS WebP Rewriting to deliver lightweight images without requiring server-level changes, successfully navigating strict security protocols that blocked standard methods.
Database-Level Emergency Recovery
Utilized WP-CLI for emergency database surgery to force a 'Safe Mode' bypass, clearing corrupted plugin options via the terminal to immediately restore site availability after a critical error loop.
Optimized Caching & Delivery
Implemented a multi-layered caching strategy including asset minification, intelligent lazy loading, and global cache purging to enhance load speed and ensure real-time propagation of fixes.
Final Results
Successfully transformed a heavy, bug-prone mobile experience into a high-performance, stable platform. The site now utilizes next-generation image formats and optimized delivery methods that work in harmony with high-security hosting environments, ensuring a fast and reliable experience for all users.
Tech Stack
Project Goal
To engineer a fully autonomous, 24/7 social media marketing agent that moves beyond simple scheduling. The system is designed to intelligently source, create, and publish content using a Hybrid Retrieval-Augmented Generation (RAG) Architecture, ensuring resilience and relevance.
Key Features
Live Intelligence Gathering
Scrapes real-time web data to discover breaking industry news and relevant topics, ensuring content is always timely.
Contextual Content Synthesis
Uses Google's Gemini to synthesize long-form articles and reports into engaging, platform-specific threaded conversations.
Generative Visuals
Autonomously prompts DALL-E 3 to create context-aware memes and promotional imagery that align with the generated text.
Self-Healing Architecture
Features a robust try-except-cache loop. If an API rate limit or connection drop occurs, the system caches the content and retries, ensuring zero data loss.
Outcome
The agent successfully runs continuously on a Linux server, autonomously generating and posting high-quality, relevant content to social media platforms like Twitter and LinkedIn. Its self-healing capabilities have proven effective in handling network and API interruptions, resulting in a truly hands-off marketing automation solution.
Tech Stack
Project Goal
To develop a fast-paced, responsive web game, demonstrating proficiency in front-end development, game logic implementation, and real-time database integration for a global leaderboard.
Key Features
Real-Time Rendering Loop
Utilized the HTML5 Canvas API and `requestAnimationFrame` for a smooth, high-performance game loop, independent of React's render cycle.
Advanced State Management
Leveraged React Hooks (`useState`, `useRef`, `useCallback`) for efficient state management and to prevent re-renders during active gameplay.
Global Leaderboard
Implemented a global leaderboard using Firestore, including secure data submission and real-time score updates.
Responsive Controls
Created a seamless user experience with intuitive controls across both desktop (keyboard) and mobile (touch) devices.
Future Expansion
This project serves as a foundation for more complex web-based games. Future plans include adding power-ups, more enemy types, and progressively difficult levels to enhance gameplay depth.
Tech Stack
Project Goal
To develop a sophisticated algorithmic trading bot that executes trades based on custom strategies written in Pine Script. This bot is designed to automate trading decisions and capitalize on market opportunities 24/7.
Key Features
Custom Strategy Development
Leverages Pine Script to create and implement complex, custom-tailored trading strategies from the ground up.
TradingView Integration
Seamlessly integrates with TradingView for real-time charting, analysis, and alert-based trade execution.
Automated Execution
The bot autonomously executes trades when market conditions meet predefined criteria in the Pine Script strategies, operating 24/7.
Robust Backtesting
Utilizes Backtesting.py and Python to rigorously test and optimize trading strategies against historical data, ensuring viability and performance before deployment.
Outcome
The resulting system provides a powerful framework for quantitative trading, enabling rapid prototyping and validation of complex strategies. It successfully automates the entire workflow from idea to execution.