Hi everyone - Nate here.
It’s been a while since the last module because I’ve been working on a lot of other projects, which I will share eventually. But, I’ll try to keep up on these core learning modules.
This module covers customization of LLMs, with a hierarchy of methods from the simplest to most complex, with a particular focus on fine-tuning. This gets at the question: When should you actually fine-tune a model versus just writing better prompts or using RAG? The module walks through all your options, from testing prompts to full fine-tuning. Key insight: fine-tuning changes behavior, not knowledge. It won't make your model "know" your organization's data - that's what RAG is for.
Also covered:
Platforms for advanced prompt design and testing
Structured outputs and tool calling
When to use RAG vs. fine-tuning
A brief overview of fine-tuning and using no-code platforms to fine-tune LLMs
LLM customization options for global health purposes
Happy learning!
