Rconsortium

Keeping LLMs in Their Lane: Focused AI for Data Science and Research

From R+AI 2025, hosted by R Consortium

Keynote

LLMs are powerful, flexible, easy-to-use… and often wrong. This is a dangerous combination, especially for data analysis and scientific research, where correctness and reproducibility are core requirements. Fortunately, it turns out that by carefully applying LLMs to narrower use cases, we can turn them into surprisingly reliable assistants that accelerate and enhance, rather than undermine, scientific work. This is not just theory—I’ll showcase working examples of seamlessly integrating LLMs into analytic workflows, helping data scientists build interactive, intelligent applications without needing to be web developers. You’ll see firsthand how keeping LLMs focused lets us leverage their “intelligence” in a way that’s practical, rigorous, and reproducible.

Bio

Joe Cheng is the CTO and first employee at Posit, PBC (formerly known as RStudio), where he helped create the RStudio IDE, Shiny web framework, and Databot agent for exploratory data analysis.

R Consortium Resources

Personal R Administration

From R/Medicine 2025

– Does the release of a new R version fill you with dread? – Are there passwords in your R code? – Do you look at the output of a failed package installation and think to yourself, “WTF?!”

If you said yes to any of those questions, then you need Personal R Administration. You’ll come away with tips, tricks, tweaks, and some hacks for building data science dev environments that you won’t be afraid to come back to in a year.

David Aja and Shannon Pileggi

E. David Aja is a Software Engineer at Posit. Before joining Posit, he worked as a data scientist in the public sector.

Shannon Pileggi (she/her) is a Lead Data Scientist at The Prostate Cancer Clinical Trials Consortium, a frequent blogger, and a member of the R-Ladies Global leadership team. She enjoys automating data wrangling and data outputs, and making both data insights and learning new material digestible.

Resources

Demystifying LLMs with Ellmer

From R/Medicine 2025

Joe Cheng, CTO, Posit Joe Cheng is the CTO of Posit, PBC. He’s the original creator of the Shiny web framework and co-creator of ellmer.

Today’s best LLMs are incredibly powerful–but you’re only scratching the surface of their capabilities if your use is limited to ChatGPT or Copilot. Accessing LLMs programmatically opens up a whole new world of possibilities, letting you integrate LLMs into your own apps, scripts, and workflows. In this workshop, we’ll cover:

A practical introduction to LLM APIs

– Configuring R to access LLMs via the ellmer package – Customizing LLM behavior using system prompts and tool calling – Creating Shiny apps with integrated chatbots – Using LLMs for natural language processing

Attendees will leave the workshop armed with ready-to-run examples, and the confidence and inspiration to run their own experiments with LLMs.

Attendees should be familiar with the basics of R and have a working R installation.

Note that to avoid any potential firewall issues, it is recommended that participants use a personal computer for this workshop.

Resources

Quarto Dashboards: from zero to publish in one hour

From R/Medicine 2025

You already analyze and summarize your data with R and Quarto. What’s next?

You can share your insights or allow others to make their own conclusions in eye-catching dashboards and straight-forward to author, design, and deploy Quarto Dashboards. With Quarto Dashboards, you can create elegant and production-ready dashboards using a variety of components, including static graphics, interactive widgets, tabular data, value boxes, text annotations, and more.

Additionally, with intelligent resizing of components, your Quarto Dashboards look great on devices of all sizes. And importantly, you can author Quarto Dashboards without leaving the comfort of your “home” – in plain text markdown with any text editor. In this one-hour demo we will build and publish a Quarto Dashboard – you can code-along or sit back and enjoy the show!

Mine Çetinkaya-Rundel, Professor of the Practice at Duke University and Developer Educator at Posit

Mine’s work focuses on innovation in statistics and data science pedagogy, with an emphasis on computing, reproducible research, student-centered learning, and open-source education as well as pedagogical approaches for enhancing retention of women and under-represented minorities in STEM.

Mine works on the OpenIntro project, whose mission is to make educational products that are free, transparent, and lower barriers to education. As part of this project she co-authored four open-source introductory statistics textbooks – latest is the 2nd edition of Introduction to Modern Statistics.

She is also a co-author on R for Data Science, the creator and maintainer of Data Science in a Box, and she teaches popular data analysis and data science with R courses on Coursera. Mine is a Fellow of the ASA and Elected Member of the ISI as well as a Waller and Hogg award winner for teaching excellence. In 2024, she was elected as Vice President of the International Association for Statistical Education (IASE).

Next Generation Shiny Apps with {bslib}

In this workshop, led by Garrick Aden-Buie, a software engineer for Shiny at Posit, you will learn how to build Shiny apps using modern user interfaces and layouts with bslib, the next generation of Shiny UI. Garrick will guide you through creating stylish and convenient dashboard layouts and components, demonstrating how bslib can effectively replace shinydashboard. Additionally, you will explore innovative techniques for deploying Shiny apps using shinylive, enabling the creation of static sites that run entirely in the user’s browser without the need for a Shiny server.

Main Sections

00:00 Introduction 11:00 Shiny 14:57 How bslib started 18:29 Theming from start to advanced 22:19 College Scorecard Data 39:24 bslib layouts 01:20:25 Advanced layouts 01:49:18 Filling Layouts 02:42:09 Details on demand 02:50:01 New inputs 02:53:15 Wrap up

More resources

R/Medicine: Quarto for Reproducible Medical Manuscripts

In this webinar, Mine Cetinkaya-Rundel, Professor of the Practice of Statistical Science at Duke University,  presents a new capability in Quarto that provides a straightforward and user-friendly approach to creating reproducible manuscripts that are publication-ready for submission to science journals. 

This new feature, Quarto manuscripts, includes the ability to produce a bundled output containing a standardized journal format, source documents, source computations, referenced resources, and execution information into a single bundle that be ingested into journal review and production processes. 

In this talk, Mine demonstrates how Quarto manuscripts work and how you can incorporate them into your current manuscript development process. She also touches on pain points in your current workflow that Quarto manuscripts help alleviate.

Main Sections

00:00 Introductions 03:05 Full complexity spectrum of reproducible of reproducible scientific projects 08:14 Leveraging Quarto for fully reproducible scientific manuscripts 16:38 A new project type: manuscript 17:12 Let’s write a manuscript 40:41 What’s next? 44:39 Thank you/Q&A

More resources

GitHub Copilot in Rstudio, it’s finally here!

Thomas Mock, PhD, Workbench Product Manager at Posit PBC.

In this webinar, part of a new quarterly R/Med seminar series, Thomas demonstrates how to set up Copilot in RStudio and then provides examples of using it to generate code by providing context and comments. Some key points covered include:

• How Copilot works by predicting the next token based on context.
• Tips for using Copilot effectively, like breaking problems down simply, specifically, and using comments.
• Examples of Copilot generating functions, tests, and repeating tasks.
• Using other tools like Chatter to ask questions when stuck.

Main Sections

00:00 Intro 01:35 What is generative AI? 04:27 What is Copilot? 08:50 Copilot in RStudio 10:15 Get started 13:01 Getting the most out of the generative loop 15:14 Simple and specific 24:29 Getting stuck? 28:53 {chattr} package 31:17 Generative AI tools with Posit Workbench and RStudio 32:43 Examples using Copilot inside RStudio 51:42 Q&A and RStudio User Guide

More Resources

BioC 2022 - Hello, Quarto!

Mine Çetinkaya-Rundel, PhD., Professor of the Practice at Duke University, Data Scientist and Professional Educator at RStudio, Inc., gives her keynote presentation at the Bioconductor Conference 2022. Dr. Cetinkaya presents Quarto, an open-source scientific and technical publishing system built on Pandoc.

Dr. Mine began her presentation by introducing Quarto and her personal experience using and teaching it to others. She then continued by giving an overview of the R Markdown ecosystem which includes packages like xaringa, Distill, Blogdown, Rmarkdown, and more. She explained the use of Quarto with these packages along with some other Quarto highlights. Afterward, Dr. Mine gave her first demo of Quarto which included setting up and sharing some handy features. Dr. Mine also gives another demonstration on publishing with Quarto and an overview of the features of collaboration. Dr. Mine then gives a second demonstration of collaborating and teaching with Quarto. To wrap up the presentation, Dr. Mine shared about reimagining open source and the work done by Openscapes and their mission for open practices accelerating data-driven solutions, and increasing diversity, equity, inclusion, and belonging in science. The presentation ended with a questions and answers session from the audience.

Main Sections

0:00 Introduction 4:29 Quarto! 6:12 Share 10:57 The R Markdown ecosystem 11:59 Quarto highlights 13:40 Demo of Quarto 21:40 Demo Quarto Publishing 22:58 Quarto rundown 25:23 Collaborate 28:43 Demo 2 34:07 Teaching with Quarto 40:39 Reimagine 44:03 Q&A and Resources

More Resources