Fine-Tuning Your AI Models: from Beginner to Advanced

This is the repository for the LinkedIn Learning course Fine-Tuning Your AI Models: from Beginner to Advanced. The full course is available from LinkedIn Learning.

Course Description

Fine-tuning lets you adapt an open-weight language model to your own task, data, and brand voice. In this course, you'll learn a complete modern fine-tuning workflow by building a support-ticket assistant for Ridgeline Outfitters, a fictional outdoor gear retailer. You'll prepare and validate training data, fine-tune a model with LoRA on a free GPU, score it against the base model, and run your finished model on your own laptop with Ollama. Along the way, you'll practice prompting, synthetic data generation, and evaluation design in hands-on exercises. By the end, you'll know when fine-tuning beats prompting alone and how to deliver a fine-tuned model with measurable business value.

Learning Objectives

• Determine when fine-tuning is the appropriate solution compared to prompting or retrieval-augmented generation (RAG).
• Prepare and validate training datasets for LLM fine-tuning, including generating synthetic examples and formatting data correctly.
• Fine-tune an open-weight language model using LoRA and modern tools such as Hugging Face, PEFT, TRL, and Unsloth.
• Evaluate fine-tuned models using baseline comparisons, structured evaluation rubrics, and LLM-as-a-judge techniques.
• Deploy and integrate a fine-tuned model into a practical business workflow.

Instructions

This repository has a folder for each chapter of the course that includes code: CH03 through CH07.

Inside each chapter folder, every lesson with code has two folders. The folder ending in _begin holds the code as it is at the beginning of the video. The folder ending in _end holds the code as it is at the end of the video. For example, CH03/03_03_begin and CH03/03_03_end go with video 03_03, the third video in chapter 3.

Every chapter folder also has its own ridgeline.py helper file and data folder, so each chapter works on its own. If you skip a challenge, the next chapter still has the files it needs. Each chapter folder has a README with links to its notebooks.

Running the Notebooks in Google Colab

You run this course's notebooks in Google Colab, which gives you a free GPU in your browser. There's nothing to install on your own machine until the final chapter.

To open the notebook for any video:

1. Open the chapter folder, then the lesson's _begin or _end folder, then the notebooks folder, and select the notebook.
2. Select the Open in Colab button at the top of the notebook. Every chapter README also has these buttons. If you don't see one, copy the notebook's GitHub URL and replace https://github.com/ with https://colab.research.google.com/github/.
3. In Colab, choose Runtime > Change runtime type, select T4 GPU, and select Save.
4. Choose File > Save a copy in Drive so your changes are saved to your own Google Drive.

Start with the GPU check notebook to confirm your setup:

Notebook	Open
Check your GPU

Colab's free tier has usage limits, and idle sessions disconnect. If you lose your GPU mid-course, wait a while and reconnect, or move on to the next chapter, which has its own copy of the files it needs.

Installing

1. To use these exercise files, you must have the following:
   • A free Google account, to run the notebooks in Google Colab
   • A current web browser
   • For the final chapter only: Ollama installed on a computer with at least 8 GB of RAM and about 5 GB of free disk space, plus Python 3
   • Optional: a free Hugging Face account, if you want to publish your fine-tuned model
2. Clone this repository into your local machine using the terminal (Mac), CMD (Windows), or a GUI tool like SourceTree. You only need a local copy for the final chapter, when you run your model with Ollama. Every other chapter runs in Colab.
3. Open CH03/notebooks/00_check_gpu.ipynb in Colab using the steps in Running the Notebooks in Google Colab and run both cells. If both print a success message, you're ready to start the course.

Repo Layout

• CH03 to CH07: one folder per chapter with code
  • XX_YY_begin and XX_YY_end: the starting and finished code for each lesson
  • ridgeline.py: shared helpers for the Ridgeline Outfitters project
  • data: the tickets, policies, and checkpoint files the chapter uses

Instructor

Kesha Williams

Kesha Williams is an AWS AI Hero and the founder of Keysoft, an enterprise AI strategy, engineering, and education consultancy.

Check out my other courses on LinkedIn Learning.