Inference Gateway
Reliable, self-healing model serving on shared GPU hardware.
The infrastructure layer that keeps our AI reliable. It lets several projects share a single GPU without colliding — serializing inference so only one model runs at a time, queueing the rest, and detecting and auto-healing degraded output before it reaches users. A single static binary with a live dashboard for queue depth, GPU health, and recent requests.
What it demonstrates: production discipline around AI — the unglamorous reliability work that separates a demo from something you can run every day.