LLM Agents for Research Using R - A Short Course

You will walk away from this course with the ability to:

• • Create agents for literature reviews, measurement identification, and research support.
  • Integrate web search and external data sources into agent workflows.
  • Coordinate multiple agents to complete complex tasks.
  • Document AI use in academic research.
  • Evaluate and verify agent outputs using research best practices and ethical considerations.

This is a hands-on course with instructor-led software demonstrations and guided exercises. These guided exercises are designed for the R programming language, so you should use a computer with a recent version of R (version 4.1.3 or later) and RStudio (version 2022.02.1+461 or later). The names of the packages needed and how to install them will be provided during the course.

You will need to set up API keys to use LLM systems hosted through cloud services. Some commercial models may require adding a small amount of credit, typically $5–10 USD, to your account (e.g., OpenAI). LLM models hosted by cloud services are typically free to use, even when an API key is required. Specific setup instructions will be provided before the course.

To follow along with the course exercises, you should have good familiarity with the use of R, including opening and executing data files and programs, as well as performing very basic data manipulation and analyses.

If you’d like to take this course but are concerned that you don’t know enough R, there are excellent online resources for learning the basics. Here are our recommendations.

This seminar is designed for researchers who are interested in using AI to automate and enhance their work. You should have:

• • Basic familiarity with R and RStudio, including the ability to run scripts and install packages.
  • An interest in applying AI tools to research tasks such as literature reviews, measurement identification, and research design.

Some prior experience using large language models in R will be helpful but is not required. For additional preparation, you may wish to take LLM Text Generation in R: Practical Prompt Engineering.

No Python experience is necessary. All Python-based agent functionality will be accessed through R.

In this seminar, you’ll learn to build custom research agents using R and APIs. If you’re looking for a less programming-focused introduction to using agents responsibly, you may be interested in AI Agent-Driven Research Workflows. If your goal is to organize file-aware agent work across multiple projects, consider Getting Work Done with AI Agents.

Module 1: Introduction to LLMs

Understanding the history of AI, basic notions of LLMs, transformers, and prompt engineering.

• • Brief history of AI
  • What are LLMs?
  • What is prompt engineering?

Module 2: Introduction to LLM agents

Understanding the agent paradigm and how it extends beyond simple prompting.

• • What are agents? Key components: LLM, tools, memory, instructions
  • Agents vs. chatbots: autonomous action and tool use
  • Overview of the agent ecosystem
  • Setting up your development environment

Module 3: Free and open-source models via NVIDIA’s GroqCloud system

Using free and open-source models via GroqCloud’s built-in web-search capabilities

• • Introduction to available models
  • Basic web search
  • Domain filtering for academic sources
  • Search query modifiers
  • Comparing predefined models with custom agent approaches

Module 4: Building custom agents with an agentic framework

Creating flexible, instruction-following agents with an agentic framework.

• • Installing and configuring the agentic framework with Groq models
  • Agent anatomy: model, name, description, tools, instructions
  • Adding tools: for web search and online data retrieval
  • Crafting effective agent instructions for research tasks
  • Model selection: when to use smaller models vs. larger models

Building specialized agents for common research tasks, such as:

• • Literature review agent: searching and summarizing academic literature
  • Measurement finder agent: locating validated scales and instruments
  • Research design agent: helping plan rigorous studies
  • Grant writing agent: finding funding opportunities and strategies

Module 5: Multi-agent teams, online data, and data agents

Coordinating multiple specialized agents for complex tasks.

• • Orchestrating multiple agents
  • Agent roles and task delegation
  • Integrating real-time data sources into research workflows
  • Combining web search with online data retrieval
  • Building agents for research
  • Handling API errors and rate limits