Modal

Modal is a serverless compute platform designed for AI and data workloads. Developers define functions and environments with Python decorators and run them on on-demand cloud GPUs/CPUs with fast cold starts and automatic scaling. It covers model inference, fine-tuning, large batch jobs, scheduled tasks, and code sandboxes, billed by actual usage. For AI engineers and teams who don't want to manage GPU clusters themselves.

• On-demand serverless GPUs
• Infrastructure defined in Python
• Fast cold starts and autoscaling