AI Resources for Researchers

Azure AI Environment provides secure, HIPAA-compliant access to advanced AI tools. Transmit data via API and work inside the platform with a paid subscription through IU.

IU’s Jetstream2 has ninety GPU nodes for large-scale computations, running on Ubuntu Linux. Ideal for projects requiring significant power for datasets that are not covered by HIPAA.

IU’s Information Technology Services can configure a HIPAA-compliant node customized to your AI hardware needs.

Pre-trained models can be downloaded and run on local hardware (e.g., see huggingface.co or lmstudio.ai). While convenient, these models may need additional hardware capabilities to suit specific research goals.

For projects requiring customized models, fine-tuning an existing LLM is an option (view a Datacamp guide on the subject). This process typically requires moderate to substantial computational resources, which can be accessed via a single Quartz node in Regenstrief’s infrastructure. However, some of the hardware may not be the most up-to-date for the latest AI needs.

Creating an LLM from the ground up is a highly computationally intensive process, often costing millions of dollars in computing power. This approach is generally reserved for those with in-depth LLM knowledge when addressing exceptional cases.