LLM Agents for Data Science and Research
Building Grounded, Tool-Using AI Agents for Literature Review & Evidence Synthesis
Rolling Basis – Course offering for new cohort begins every 5 weeks
Live via Zoom
LLM Agents for Data Science and Research (8-Day Weekend Program)
Duration: 8 days (4 weeks)
Schedule: Saturday/Sunday, 4:30-5:30 PM ET (New York Time)
Registration closes: October 31, 2025
Format: Live via Zoom
Payment Options: Bank Transfer, Venmo, or Zelle accepted.
Course Fee: $250 (limited scholarships and variable pricing available based on Country of participant)
About the Training
This course provides a hands-on introduction to building AI research agents in R, using the ellmer and ragnar packages to connect large language models to real research workflows.
Participants begin with foundational LLM interactions and structured data extraction, progress through tool-calling and retrieval-augmented generation, and conclude with multi-step, self-checking research agents.
The course emphasizes grounding, citation, and verification — building agents that retrieve real evidence and compute real statistics rather than generating plausible-sounding guesses — for use in literature review and evidence synthesis contexts.
- Instructor-led, live Zoom sessions
- Real literature corpora & structured research datasets
- Chat calls → tool-calling → RAG → multi-step agent progression
- Strong focus on grounding, citation & factuality
- Capstone: a self-checking, cited research-assistant agent
Who Should Register
Researchers wanting to automate literature review and extraction; R users comfortable with dplyr who are new to LLM tooling; Analysts curious about agents beyond a chatbot demo.
- Basic R and statistics familiarity
- No prior ML experience required
- Conceptual, not purely mathematical approach
Learning Objectives
- Interact with LLMs programmatically in R using the ellmer package, including structured data extraction from unstructured text into tidy data frames.
- Design and register custom tools that let an LLM call real R functions, building agents that compute real answers instead of generating plausible guesses.
- Build a retrieval-augmented generation (RAG) pipeline - chunking, embedding, and retrieving source documents - to ground LLM answers in real evidence with citations.
- Chain multiple agent capabilities (retrieval, extraction, computation) into a multi-step research workflow with self-correction and factuality checks.
- Produce a reproducible, cited research report generated by an autonomous agent, suitable for a literature review or evidence synthesis task.
Apply machine learning workflows in R for real-world datasets.
Build, tune, and evaluate predictive and unsupervised models.
Implement feature engineering and interpret model outputs.
Communicate findings using reproducible R Markdown or Shiny dashboards.
Design and present an end-to-end ML project.
Weekly Breakdown
Each week includes:
🎓 Lecture slides (PowerPoint)
💻 R lab exercises (ellmer + ragnar)
📊 Synthetic research datasets (paper abstracts & literature corpus)
Computing & Tools
All analyses are done in R using packages like ellmer, ragnar, dplyr, and httr2 — the official tidyverse toolkit for connecting R to OpenAI, Claude, Gemini, and local LLMs.
All example files and syntax are downloadable from the course portal.