Work with AI

Use large language models as programmable tools for extraction, analysis, retrieval, and application development

The AI · LLMs track: four lessons plus a live capstone, from calling an LLM as a function to a deployed retrieval-augmented chatbot.

Track description

This track treats large language models as programmable tools, not just chat windows. You’ll call an LLM from R, force its output into a typed shape, give it tools it can decide to call, put it inside a Shiny app, and finally build a real retrieval-augmented generation (RAG) pipeline so it answers from actual retrieved text instead of memory – ending with a live, deployed capstone built the same way.

Who this is for

Anyone who wants to use LLMs as part of an actual R workflow – extraction, retrieval, tool calling, or an app – rather than as a standalone chatbot.

Prerequisites

• Minimum: comfort with R and the tidyverse pipe (|>), as taught in the Build with R track’s first two lessons. This track assumes that, and does not re-teach it.
• Helpful: having built a Shiny app before (Make it shine) makes lesson 2 land faster, but is not required.
• No prior AI/ML background needed – every concept (embeddings, tool calling, retrieval) is introduced from scratch.
• No API key is required to read any lesson: LLM code chunks are shown with real captured output rather than executed live.

Lessons

1 Talking to machines Call an LLM from R, extract structured data from text, and give the model callable tools. Intermediate · 45 min 2 Chat with your data Embed a streaming chat UI and natural-language data filtering inside a Shiny app. Intermediate · 45 min 3 Grounded in truth Build a full RAG pipeline: crawl, chunk, embed, store and retrieve real text for an LLM to answer from. Advanced · 60 min 4 Agentic coding Understand what an agentic coding tool actually does under the hood, and the vocabulary around it. Beginner · 45 min ★ Ask R4DEV — capstone See the RAG pipeline running live, answering questions about R4DEV from its own retrieved content. Intermediate · 15 min

Estimated total time: ~3.5 hours across 4 lessons plus the capstone (durations are rounded active-work estimates, not automated reading time).

NoteAbout the capstone

“Ask R4DEV” is a real, deployed chatbot embedded at the end of Grounded in truth – it is not a separate lesson to work through, but a live demonstration of the pipeline that lesson teaches, built with the same functions. Open it after finishing lesson 3.

Start with lesson 1: Talking to machines →