How it works

A private AI service, made from memory your cluster isn't using.

Cleanroom turns your existing research cluster into a private, campus-wide AI service. It runs AI models in GPU memory that research jobs leave unused, and gives that memory back the moment research needs it.

  1. Step 1

    Research jobs run as usual

    Research jobs use your cluster's GPUs just as they do today. Many of them keep a GPU's compute busy while using only a small part of its memory, leaving the rest empty.

    Research jobCleanroom AI modelUnused memory
  2. Step 2

    Cleanroom places AI models in the unused memory

    Cleanroom puts AI models into the memory those jobs leave empty. Together, that space becomes a private AI service for your whole campus, running on GPUs you already own.

  3. Step 3

    Research always comes first

    If a research job needs its memory back, Cleanroom gives it up immediately. Your cluster exists for research, and Cleanroom is built around that.

    Research jobCleanroom AI modelUnused memory

Illustrations are simplified for clarity.

For researchers

What researchers get

Works with existing tools

One secure AI service that works with the tools and code researchers already use.

Campus login

Researchers sign in with their campus credentials, and usage can be billed to grants or departments.

One service for everyone

A single shared service handles many users at once, so researchers don't each have to request cluster time to run their own model.

Questions

Common questions

Do we need to buy new hardware?

No. Cleanroom runs on the GPUs your institution already owns.

Will it slow down research jobs?

Research jobs get their memory back immediately whenever they need it. To measure any effect on performance, we're running a controlled experiment with research software running alongside AI models. It tracks slowdown (our target is under 5%), response time, and how safely memory is handed back.

Which AI models can we use?

Your institution decides. That includes open models whose weights you keep, which supports reproducible research.

Does our data leave campus?

Only if you allow it. You set rules for each type of data, so sensitive data stays on campus. For tasks that need the most capable commercial models, Cleanroom can send requests to the cloud wherever your institution permits it.

Do researchers have to change their code?

Cleanroom provides one secure AI service that is compatible with the tools and code researchers already use.

Curious how this would work on your cluster?

We're looking for research computing teams to work with. In a first conversation, we'll compare notes on your cluster and what your researchers need from AI.