Modal Installation Hello World — Run a Function on GPU Custom Container Images Model Serving with Web Endpoints Volumes for Persistent Storage Parallel Map (Batch Processing) Scheduled Jobs (Cron) Secrets Management Key Concepts - Scales to zero : No cost when idle; containers spin up on demand - GPU selection : , , , — pick by workload - : Runs once when container starts — ideal for loading models - Volumes : Persistent storage shared across function calls for model caching - : Fan out work across many containers in parallel - Web endpoints : Auto-generated HTTPS URLs with authentication - f…