Skip to main content

Getting Started with AIR

The ASU AI Research Acceleration Platform (ASU AIR) is a collection of hardware and services, hosted at ASU by Research Computing, which enable AI research at ASU. This platform is available to all individuals at ASU, and here we provide information to enable use of the platform.

What is AIR?​

ASU received a donation from Intel in Fall 2025 consisting of Intel Gaudi2 HPUs. These accelerator cards are specialized to ensure high performance for AI and ML applications. The AIR Platform collects these cards into a single platform, which can be accessed by anyone at ASU.

What services are provided through AIR?​

The AIR Platform provides multiple methods of use, and more are being added over time. Highlights include:

  • API key access, enabling external applications to leverage AIR
  • Tooling to enable agent usage
  • Chatbot interfaces by way of CreateAI Builder and Open WebUI
  • Kubernetes access, enabling:
    • Jupyter Notebooks
    • Pre-packaged bioinformatics programs
    • Custom LLM deployment
    • Custom research applications

How can I access AIR?​

The first step in using AIR is signing up. All registrations are handled via the Voyager RC Resource Manager. For full instructions on registration, please refer to this documentation, requesting a non-LLM account.

What can I do with my access?​

info

The ASU SSLVPN is required to connect to the AIR services through the web. On-campus networks still require VPN.

Note that creating an API key will require VPN access, while using it does not.

With an account, you have full access to the suite of AIR Platform services. To start, consider using Open WebUI for a familiar chat interface.

Model details​

Through the AI and LLMs tab on Voyager, you can find details on our currently installed models, along with information for each one. This information is useful when selecting models to use with other tools.

API Keys​

The Keys tab on Voyager allows you to generate and track information on API keys. All keys generated through Voyager for the AIR Platform are OpenAI API compatible, so you can use them anywhere you would expect to be able to leverage other AI platforms. These are perfect for applications that are hosted elsewhere and rely on AI as only a part of their overall function, or to provide the AI component in an agentic environment such as OpenCode.

For information on getting your own API keys, refer to our tutorial. To limit where a key can be used from, or to see where it was last used, see Keys and allowed networks.

Running tools directly on AIR​

Using the Workflows and Apps tab on Voyager provides you access to both preconfigured and custom applications you can launch or create on the AIR Platform, and are intended for when an application is concerned with running directly on the system rather than placing external calls to the API. Examples include: using it for complex applications such as bioinformatics programs; backing research on ML and AI methods on a Jupyter Notebook; and hosting your own LLM for research purposes via vLLM.

Where can I learn more?​

The services available through the AIR Platform are evolving with time, and you can always access them via Voyager. While we did not discuss them in detail here, we maintain documentation on our current services.

The following is a living list of current tutorials and resources provided to enable your use of the AIR Platform: