Engineering evidence & results
I want this lab to be more than a visually interesting collection of dashboards. A useful engineering portfolio shows how a system was designed, what was actually built, and what happened when it was tested.
Existing work and direct evidence
| Area | Evidence |
|---|---|
| Thermostat | Complete public build article, published October 22, 2026, documenting the hardware, control algorithm and successful integration. |
| Service organization | Screenshot of Buffy's private dashboard, plus the searchable software inventory. |
| AI research platform | Platform documentation and the described MLOps toolchain; experiment-level measurements are the next step. |
| Observability | Monitoring architecture and the service inventory; a dated incident walkthrough is planned. |
What I intend to publish next
- An RTX experiment with model revision, input, GPU memory use and measured inference latency.
- A reproducible ML workflow with DVC revision, Prefect run steps, MinIO artifacts and logged outputs.
- A monitoring case study describing one actual service failure, investigation and recovery.
- A thermostat follow-up with longer-duration room-temperature trends and control outcomes.
- A working public inference interface, isolated from internal management services.
Measured results will include a date and test setup so others can interpret them. I don't use estimated results as benchmark evidence.
Links
Start with the homelab overview, thermostat story, or MLOps workflow.