Building My AI Cluster
· One min read
I wanted a simple way to experiment with local inference, distributed services, and observability without building a lot of custom infrastructure first.
Goals
- Use hardware I had laying around
- Server the model on my local network
- Use any AI harness I wanted
Current stack
- llama.cpp (build from source)
- Pi.dev (AI harness)
- 3 Raspberry Pi 5 + 3 PoE Hats + 4-Port 2Gbps Network Switch
- 1 2TB SATA Drive
Next steps
Come back for the full blog post