Generative AI

Large Language Models (LLMs) are one of the fastest ways to get started with the Hackathon APIs: it is often quicker to ask one for working FHIR or DICOMweb code than to read through example code. A free account on a service such as ChatGPT, Claude or Gemini is generally enough for hackathon use.

NVIDIA's Build Page

NVIDIA hosts a number of open-source Generative AI models at build.nvidia.com. The models range from Large Language Models (e.g. llama) to visual models (generate images based on textual prompts) and many other categories. Additionally, the website allows you to try out the different models without having to login. It makes for a nice way to get a feel for what each model is capable of, but also to try these models before you invest time/money into hosting your own copy of them.

Keep in mind...

LLM-generated code is only a starting point and may not work as-is against the Hackathon server for a variety of reasons. Some basic troubleshooting and minor tweaks might be necessary. Don't be afraid to ask for help if you get stuck! And please share on Slack what you're using LLMs for and how useful you found them.

Acknowledgements

Earlier hackathons, when free LLM accounts were far more limited, offered free GPT access through a proxy generously provided by MD.ai and Dr. George Shih. Thank you to them, and to Dr. Howard Chen and Dr. Stephanie Hou for documenting, testing and reviewing it.