Gabriel Cosi

Software Engineer

Systems Operational

Running LLMs at home

Building an on-demand inference box.

LLM Usage

Tokens (7d)7,333,211 tokens

Top Model (7d)qwen3.6-35b-a3b

Cluster Metrics

CPU Load33.1%

Memory42.7%

Active Pods200

Uptime31d

SLA (7d)99.98%

Global P99 (7d)80 ms