Architecting the
AI-First Ecosystem
I build high-performance infrastructure for the AI era. Specializing in LLM-integrated Web Apps, Offline-Capable Desktop Tools, and Resilient Backend Systems that power the next generation of automation.
Engineering Showcase
Explore the suite of tools built for efficiency, automation, and scalable intelligence.
The Ghostwriter
Style-Matching Auto-Complete. Instead of generic AI writing, this add-in uses a lightweight local model/LoRA trained on your Sent Items to mimic your personal voice perfectly.
Private Attachment Chat
Chat with your PDF, Excel, or Word attachments locally. Extract data and ask questions about sensitive documents without ever uploading them to the cloud.
Thread Archaeologist
Transforms massive email chains into a structured chronological timeline of decisions. Identifies temporal markers and decision verbs to reconstruct event flow.
Self-sustaining content generation pipeline powered by Firebase Cloud Functions and Gemini 2.0. Orchestrates daily research loops via serverless architecture to document AGI acceleration.
The Ghostwriter
The Ghostwriter redefines email composition by learning your unique voice. Unlike generic AI tools that produce robotic text, this add-in analyzes your 'Sent Items' to understand your tone, vocabulary, and sentence structure. It runs entirely locally on your machine using a fine-tuned LoRA adapter, ensuring your data never leaves your device while providing suggestions that sound essentially like you.
Key Capabilities
- Personalized Voice cloning based on your email history
- Local-only processing for maximum privacy
- Real-time sentence completion and drafting
- Seamless integration with Outlook composer
Private Attachment Chat
Stop uploading sensitive contracts and financial reports to public clouds. Private Attachment Chat creates a secure, local RAG (Retrieval-Augmented Generation) pipeline directly within Outlook. Drag and drop any document to instantly query it—extract figures, summarize clauses, or find specific dates—all without the bits ever leaving your hard drive.
Key Capabilities
- Support for PDF, DOCX, XLSX, and TXT files
- Instant answers with citation links to the source text
- Zero latency local processing
- Enterprise-grade privacy compliance
Thread Archaeologist
Thread Archaeologist saves you hours of digging through 'Re: Re: Fwd: Re:' chains. It uses advanced NLP to parse temporal markers and actionable verbs, reconstructing the entire history of a conversation into a clean, linear timeline. See exactly when decisions were made, who made them, and how the conversation evolved, all at a glance.
Key Capabilities
- Visual timeline generation for nested threads
- Automatic detection of decision points and action items
- Participant contribution tracking
- One-click export of conversation summaries
Autonomous R&D Engine
This project represents a fully autonomous publishing infrastructure built on Firebase Cloud Functions v2. The system runs on a scheduled trigger (6 PM UTC daily), executing serverless Node.js functions that interface with Google's Gemini 2.0 Flash model. The architecture handles stochastic topic selection from a high-signal vector space, enforces strict prompt engineering constraints for tone and safety, and persists generated content to Firestore. The frontend reads from the database in real-time, creating a zero-maintenance content pipeline.
Key Capabilities
- Serverless daily content orchestration via Cloud Functions
- Integration with Google Gemini 2.0 Flash Experimental
- Real-time Firestore database synchronization
- Cloud Scheduler for automated execution
- Client-side rendering with Firebase SDK