LLM Qualitative Coding

Training materials for learning to use Large Language Models for qualitative research coding and analysis.

What is this?

A hands-on training repository that teaches:

• How to make API calls to LLMs from Python
• What embeddings are and how to use them for semantic search
• How to apply LLM capabilities to qualitative coding workflows
• Practical techniques: theme extraction, direct coding, inductive clustering

Who is this for?

Research staff who want to:

• Use LLMs programmatically for qualitative analysis
• Work with interview transcripts and qualitative data
• Build AI-augmented coding and analysis tools
• Understand embeddings and semantic similarity

No prior LLM experience required. Basic Python knowledge helpful.

Repository structure

llm-quali-coding/
├── docs/                    # Session guides with step-by-step instructions
│   ├── session_01_setup.md
│   ├── session_02_embeddings_rag.md
│   └── session_03_quali_coding.md
├── examples/                # Runnable Python scripts
│   ├── 01_test_connection.py
│   ├── 02_translate_transcript.py
│   ├── 03_create_embeddings.py
│   ├── 04_relevance_filtering.py
│   ├── 05_theme_classification_embeddings.py
│   ├── 06_extract_themes_llm.py
│   ├── 07_direct_coding_llm.py
│   ├── 08_nonverbal_coding_llm.py
│   └── 09_inductive_clustering.py
├── src/                     # Reusable code modules
│   ├── chunking.py
│   ├── coding.py
│   ├── embeddings.py
│   ├── llm_tasks.py
│   ├── openai_client.py
│   └── similarity.py
└── data/                    # Sample data for exercises
    ├── sample_transcripts/
    └── themes/

Training sessions

Day 1: Foundations

1. Session 01 (1 hour): Local setup and your first LLM API call
2. Session 02 (1 hour): Introduction to embeddings and RAG

Day 2: Qualitative Coding Applications

1. Session 03 (2 hours): Qualitative coding techniques with LLMs

Getting started

1. Clone this repository
2. Follow the setup instructions in docs/session_01_setup.md
3. Complete sessions in order

Each session builds on the previous one.

Requirements

• Python 3.11+
• OpenAI API key
• Code editor (VS Code or Positron recommended)
• Basic command line familiarity

Support

• Session guides contain detailed instructions and troubleshooting
• Example scripts include inline comments
• Common issues documented in each session guide

Learning outcomes

By the end of this training, you will:

• Have a working local LLM development environment
• Understand how to use LLM APIs programmatically
• Know what embeddings are and when to use them for qualitative analysis
• Be able to apply multiple LLM-based coding techniques to qualitative data
• Have practical experience with theme extraction, classification, and clustering

---

Start here: docs/session_01_setup.md