G.H.O.S.T.

Guided Hub for Optimal Smart Technology. What started as frustration with disconnected smart devices became a full-blown obsession with building something better.

The Story

I got tired of having six different apps to control six different devices, none of which talked to each other. The "smart" home felt pretty dumb. So I built G.H.O.S.T. — a Raspberry Pi 5 running a local LLM that actually understands context and can orchestrate everything from one interface.

The first version was held together with duct tape and prayers. The current version... is held together with slightly better code and fewer prayers. It's been a crash course in API design, local AI deployment, and learning that "it works on my machine" is not a deployment strategy.

What I Learned

• Local LLMs are more capable than I expected — and more finicky than I hoped
• Good API design is invisible; bad API design haunts your dreams
• Building for yourself is different than building for users (even when you're the only user)
• Documentation isn't optional — it's how future-you survives

Key Features

• Engineered a local LLM with a custom vector-based SQL database and embedding model for semantic memory retrieval
• Designed and tuned persona-based interaction profiles to improve usability, responsiveness, and engagement
• Integrated open-source APIs and custom scripts to enable unified, event-driven automation routines
• Operationalized G.H.O.S.T. as a daily productivity interface, improving deliverables by 12%