Modal Serverless Gpu

This gives Claude the ability to spin up serverless Python functions on Modal with GPU access, which is genuinely useful if you're deploying ML models or running batch jobs that need real compute. The skill covers the full workflow: defining container images with dependencies, requesting specific GPU types (T4 through H100), autoscaling across thousands of containers, and deploying web endpoints. Modal's sweet spot is when you need GPU inference or parallel processing but don't want to manage infrastructure. You get $30/month in free credits to start. The documentation here is thorough on images, functions, volumes, and scheduling, so Claude should be able to handle everything from one-off training runs to production API deployments.