AI Workloads on metal-stack: What Runs on GPUs Today
What metal-stack can and cannot do for GPU-accelerated AI workloads today - and where we see the gaps.

Markus Fensterer
Infrastructure
View All TagsWhat metal-stack can and cannot do for GPU-accelerated AI workloads today - and where we see the gaps.

Markus Fensterer
After finally finishing my bachelor-thesis on this topic, I'd like to give you some insights on how we built a functional metal-stack installation in a 10-inch mini-rack

Quirin Vetterl
Our lead engineer [@majst01](https://github.com/majst01) lately had the opportunity to get interviewed by Gregor Schumacher, a German blogger and author of [cloudahead.de](https://www.cloudahead.de/).

Gerrit Schwerthelm
Early in 2018 [FI-TS](https://f-i-ts.de/) decided to build a Kubernetes-as-a-Service solution for financial institutions. As with every regulated industry, there are digital sovereignty concerns regarding the control and access of data and compute. Hence, going to the public cloud isn't feasible for every workload. So, we got the task to find an enterprise ready stack for our own datacenter, while also introducing the cloud native digital competency standards.
Christian Brunner