I build AI agents and intelligent software systems that can understand context, use tools, interact with computers, and turn natural-language intent into real actions.
My current focus is moving beyond chat interfaces toward practical agentic systems — combining LLMs with voice, computer use, browser automation, memory, APIs, and reliable action execution.
A personal Windows AI agent designed to operate as a real desktop assistant rather than just a chatbot.
Ruby combines Gemini-powered reasoning, voice interaction, screen awareness, browser automation, computer control, memory, and multi-step task execution into one local assistant.
What I'm exploring:
- 🧠 LLM-powered reasoning and agent workflows
- 🎙️ Voice input, TTS, and wake-word interaction
- 🖥️ Screen awareness and computer-use actions
- 🌐 Browser automation with Playwright/Chromium
- 🖱️ Mouse and keyboard automation
- 🧩 Tool/action execution for real-world tasks
- 🧠 Persistent memory and contextual preferences
- ⚙️ Multi-step commands and task orchestration
- 🔐 Local-first control with explicit configuration
Goal: build AI systems that don't just answer questions — they can understand, decide, use tools, and execute.
I am especially interested in the engineering layer behind useful AI agents:
- Agent architecture & orchestration
- Tool calling and action execution
- Computer-use agents
- Browser automation
- Voice-first assistants
- Memory and contextual state
- LLM API integration
- Multi-step task planning
- AI-assisted software engineering
- Reliable automation and guardrails
AI agents that can:
Perceive → Reason → Plan → Use Tools → Act → Remember → Improve
I'm interested in turning that loop into software that is genuinely useful in everyday workflows — not just impressive demos.
Build less software that only responds. Build more software that can act.