Customizing Azure OpenAI Models with Fine-tuning – AI Saturday Houston 2025

[TallesValiatti]

Introduction

Hi everyone 👋

On March 22, I had the opportunity to take part in AI Saturday Houston 2025, held at the Microsoft Office, where I shared insights on “How to customize your Azure OpenAI models with Fine-tuning.”

In my session, I covered topics such as:
How fine-tuning works in Azure OpenAI

• Practical scenarios where custom models add value
• Steps to prepare and upload training datasets
• Running fine-tuning jobs in Azure
• Evaluating and deploying tuned models in production

Here’s the blog post with more details:
👉 Customizing Azure OpenAI models with Fine-tuning

Features and Screenshots

Technical Details

Azure AI Foundry, Azure OpenAI and .NET (C#)

Challenges and Solutions

One of the key challenges with fine-tuning is balancing customization with cost and maintainability. During the talk, I demonstrated how to structure training data effectively and when to use fine-tuning versus embeddings or prompt engineering.

My goal with this session (and the article) is to help students, professionals, and teams understand how to make the most of Azure OpenAI fine-tuning in real projects.