Welcome to Buffy's engineering notebook
Buffy is my self-hosted Linux workstation and the center of my personal engineering lab. I use it to learn how real systems behave by building, maintaining, and connecting them—from local AI experiments to room-aware home automation.
The hardware is an HP Z440 workstation with an NVIDIA RTX 4060 Ti (16 GB) running Ubuntu. The real value lies in the software ecosystem and the experiments it supports.
Find your way around
| Explore | You'll find |
|---|---|
| Architecture | How the workstation, containers, monitoring, and physical devices relate |
| AI & ML research platform | Local GPU workloads, experiment workflows, and MLOps components |
| MLOps workflow | Dataset versioning, orchestration, artifacts, and experimental evidence |
| Smart thermostat | A complete build story: wireless sensors, Home Assistant, Z-Wave and decision logic |
| Observability platform | Measuring the behavior of containers and the Linux host |
| Self-hosted applications | How the software ecosystem is organized |
| RTX demo | How a local model can power an independent public demonstration |
| Access & reliability | The boundaries between a public showcase and private infrastructure |
| Engineering evidence | Project artifacts and measurements that make the work reproducible |
Why a homelab?
Because there is a difference between reading about a system and running one. A homelab gives me somewhere to deploy software, examine its behavior, make mistakes, and improve it without relying entirely on managed services.
My projects fall into four areas:
- GPU experiments and applied AI: local models, vision tasks, notebooks and supporting research tools.
- Self-hosted infrastructure: Docker Compose services, databases, storage and development workspaces.
- Observability: system metrics, logs, alerts, uptime checks and troubleshooting.
- Intelligent automation: ESP8266 environmental sensors, Home Assistant and a Z-Wave thermostat.
Try an experiment
The thermostat explorer works directly in the browser. It illustrates the weighted-temperature controller described in my published build story.
The RTX AI playground is the next live connection: the application code is prepared, and public inference will become available when the GPU gateway has been configured and tested.
This site grows as the homelab grows. I share the reasoning, implementation, and limitations rather than simply listing installed software.