We were the team getting paged at 3am

Why we built it

In 2023 we were running clinical summarisation on twelve GPUs in a basement in Newark. Everything available assumed you would rent capacity from someone else. Nothing assumed you already owned the hardware and legally could not move the data off it.

So we wrote a scheduler. Then a router, because the scheduler was not enough. Two other hospital systems asked to use it, and at that point it stopped being internal tooling.

We are eleven people, funded by revenue and one seed round, and we intend to stay small enough that the person who answers your support ticket wrote the code.