Co-founder & CTO — GPU Compute Exchange
- Location
- United Kingdom
capacity from many suppliers appears to buyers as one cloud behind one API. Finish and launch managed services on partner hardware: managed Slurm and Kubernetes, storage, support and SLAs. Build the outcomes layer: serverless inference sold per token, with one scheduler across GPU generations and safe sharing of each … familiar, hyperscaler‐style workflows, e.g. an 8xH100 machine running in about 90 seconds and about two minutes from sign‐up to a running instance. Kubernetes‐based control plane: adapted to orchestrate virtual machines, with high availability and upgrades that avoid coordinated customer downtime. Own networking and storage: multi‐tenant networking ...