Generative AI on Microsoft Azure

Chapter 3. Model Selection and Prompt Engineering Best Practices

Selecting the right GenAI model is a foundational step in building effective AI applications. In this chapter, we’ll explore the key factors that influence model selection, ranging from performance benchmarks to deployment considerations, and how to navigate these choices using the Microsoft Foundry Model Catalog. Once a model is selected, its performance depends heavily on how you interact with it. That’s why we’ll also dive into prompt engineering best practices across different model families, and show how you can apply these techniques directly within Microsoft Foundry to get the most out of your GenAI solutions.

Considerations for Model Selection

The process of model selection requires a structured approach to ensure alignment with your specific needs and use case. It begins by clearly articulating the specific task or problem you aim to solve. From there, it is essential to identify key features required for this task, such as multilanguage support, input methods, or programmability. A well-defined use case should include the following factors:

Modalities

Determine whether the model needs a specific input structure and/or generate text, images, audio, video, or a combination of these (multimodal). For example, a chatbot may only require text generation, while an AI-powered design assistant may need text-to-image capabilities.

Customizability and personalization

Fine-tuning allows organizations to tailor ...