Promising AI

Modal: infrastructure to take AI projects beyond the local computer

Modal offers cloud infrastructure for running AI models and tasks, with resources defined through code.

A laptop connected to processing modules that receive audio symbols and produce text cards.
Conceptual illustration of AI workloads running on cloud infrastructure. · Credit: Imagem gerada com inteligência artificial.

The prototype works on its developer's computer. The difficulty starts when other people want to use it at the same time, the files get larger, or processing requires a more powerful machine. That transition from experiment to service is Modal's territory.

Modal offers cloud infrastructure for teams that develop and run AI applications. It handles part of the computing environment needed to run models, process tasks, and serve requests.

Infrastructure is defined through code

The official introduction explains that Modal packages code and runs it in the cloud. Python is the main language for building applications and functions on the platform.

A function is a piece of a program that performs a defined task. The developer specifies resources and dependencies, including when a GPU is needed—a type of processor used in many AI calculations.

The uses described include running models to generate responses, processing batch jobs, and training or fine-tuning models. The platform also offers isolated environments for code execution.

What might this look like in a product?

Consider an illustrative example: a team creates a tool that turns audio into text. A small file works well in the prototype, but the service starts receiving several files at once.

The team could separate the processing into tasks executed in the cloud and track the result of each one. They would still need to decide how to receive the files, handle failures, and present the text to the user.

Modal provides the execution infrastructure for that work. Choosing the model and ensuring the quality of the experience remain the product's responsibilities.

Adjusting capacity also requires choices

The scaling documentation explains that functions can gain or lose execution environments according to demand. There are settings to limit the number of environments and keep capacity ready.

These options involve a practical trade-off. Keeping resources available can reduce the wait before a task starts, but increases cost. Shutting down idle resources can save money, while the next use may need to wait for initialization.

Why follow Modal?

The appeal lies in making computing capacity easier to access without requiring every team to build its entire infrastructure operation from the start. This can create room to work on the function the user actually wants.

A useful evaluation starts with a representative task and defined usage limits. It then measures response time, cost per processing task, and behavior when something fails.

It is also worth testing the arrival of several requests at once and checking where files and results are stored. For those building AI products, Modal is promising because it addresses a crucial stage: making a good idea work beyond the development computer.

Official sources consulted on October 21, 2026.

EN

Ernani Nascimento

Developer, entrepreneur and business administrator. I build software and share what I learn along the way.

Meet the author ↗

Keep exploring.

FROM IDEA TO PRACTICE

Have something in mind?
Let’s build it.

Let’s talk ↗