Speaker
Description
While standard LLMs are powerful for general coding, they often fall short when working with specialized or internal R packages. This usually comes down to a knowledge gap - public models simply do not have access to private packages or the most recent documentation. To solve this, we propose a shift away from a single, all-purpose assistant toward a decentralized network of agents, where each R package is supported by its own dedicated expert.
This talk explores the architecture behind this system. By using a multi-agent framework, specialized agents are created on demand and grounded in real-time documentation to ensure technical accuracy. The session covers how stateful transitions are managed between these agents and the specific methods used to help them write, debug, and explain complex clinical code without losing reliability.
Additionally, the talk examines the technical side of managing a large network of experts as a scalable alternative to relying on one massive model. Key topics include the automated ingestion of R help files for grounding and the deployment of a backend service that supports multiple interfaces, from interactive chat to MCP servers. This session offers a practical framework for building AI tools in technical fields where precision and scalability are essential.
If you used AI tools or services to support the preparation of this submission, please state the name and reason for using each of them.
Yes - this presentation is about application of AI and R.
Additional Material or Paper
R+AI 2025: https://rconsortium.github.io/RplusAI_website/Abstracts.html#building-a-better-r-ai-assistant-a-multi-agent-approach
R/Pharma 2025
| Keywords: Please list up to 5 keywords to help us find the right session for your contribution. | AI, programming |
|---|---|
| Virtual Option | This submission is for onsite presentation primarily, but I would also like it to be considered for pre-recorded virtual presentation if I don't get an onsite slot |
| Video Recording | Video sharing is fine |
| The author(s) agree(s) to take responsibility and be accountable for the contents of the submission and is/are authorized to present it. | Confirm |