tenantplaneBook a demo Shared GPUs should be
simple to use.
Tenantplane builds the console for GPU clusters on Kubernetes: one place for every team to run jobs, serve models and see where the GPUs go.
Why we built it.
GPU scheduling on Kubernetes has converged on open standards. Kueue decides when work is admitted, the default scheduler places it, and llm-d routes inference. The hard part left is the experience: researchers stuck on unexplained Pending jobs, quotas kept in spreadsheets, and every model needing its own setup before anyone can call it.
Tenantplane is the layer people actually touch. Every status is one plain sentence about what is happening and what would change it.
Built on open source.
Underneath are the open-source NVIDIA Run:ai control plane for projects, access control and workload APIs, Kueue for quotas and admission, and llm-d with vLLM for serving. Tenantplane doesn’t replace the Kubernetes scheduler or add a proprietary one, so you keep every upstream improvement.
Installed with you.
We set Tenantplane up on your cluster alongside your existing workloads, configure team quotas with you, and hand it over. We also offer deployment services for teams bringing their own models into production.
Get in touch.
Zaid Fakhruddin leads Tenantplane. Write to zaidfakhruddin2@gmail.com to book a walkthrough, or try the interactive demo first.
Let your teams share GPUs
without the help desk.
Book a 30-minute walkthrough of your GPUs, teams and workloads.