This session explores the integration of AI agents into preventative healthcare with AutoGen, an open-source agentic AI framework. The solution employs a multi-agent system connected to open-source LLMs such as Llama-3.3-70B-Instruct and DeepSeek-R1-Distill-Llama-70B, hosted with Kubernetes infrastructure, and built on top of the Open Platform for Enterprise AI (OPEA). It streamlines patient data filtering and generates personalized communication. Attendees will gain insight into the architecture of the agentic framework, the OpenAI compatible API and models hosted on cloud servers, and the front-end elements with Streamlit and Hugging Face Spaces. The session will very practically cover how these technologies work together to define screening criteria, filter patient data, and automate the generation of outreach emails.

Additional Resources: The materials referenced in the session can all be accessed in the links below.

• Healthcare maintenance app and code: https://huggingface.co/spaces/Intel/preventative_healthcare
• Autogen GitHub: https://github.com/microsoft/autogen
• OpenAI best practices for prompt engineering: https://help.openai.com/en/articles/6654000-best-practices-for-prompt-engineering-with-the-openai-api
• Creating a new Hugging Face Space: https://huggingface.co/new-space
• Mike Lynch’s Medium article on preventative healthcare maintenance: https://medium.com/@micklynch_6905/hospitalgpt-managing-a-patient-population-with-autogen-powered-by-gpt-4-mixtral-8x7b-ef9f54f275f1
• Denvr Dataworks: https://www.denvrdata.com/intel
• U.S. Preventive Services Task Force for preventive care: https://www.uspreventiveservicestaskforce.org/uspstf/recommendation-topics/uspstf-a-and-b-recommendations
• My LinkedIn: https://www.linkedin.com/in/bconsolvo
• Connect with me on Intel’s DevHub Developer Discord: https://discord.gg/kfJ3NKEw5t