Welcome to Metaflow
A human-centric framework designed to help scientists and engineers build and manage real-life AI and ML systems
What is Metaflow?
Metaflow is a framework that streamlines the entire AI/ML development lifecycle—from rapid prototyping in notebooks to reliable, maintainable production deployments. Originally developed at Netflix and now supported by Outerbounds, Metaflow powers thousands of AI and ML systems across diverse companies including Amazon, DoorDash, Dyson, Goldman Sachs, and many others. At Netflix alone, Metaflow supports over 3,000 AI and ML projects, executes hundreds of millions of data-intensive high-performance compute jobs, and manages tens of petabytes of models and artifacts.Why Metaflow?
Rapid Prototyping
Start fast with notebook support, built-in experiment tracking, and local development
Effortless Scaling
Scale horizontally and vertically in your cloud with CPUs and GPUs
Production Ready
Deploy to production orchestrators with one-click and reactive event triggering
Multi-Cloud
Run on AWS, Azure, GCP, or Kubernetes with the same code
Key Features
Metaflow provides a simple and friendly Pythonic API that covers the foundational needs of AI and ML systems:Local Development
- Rapid prototyping with Python
- Built-in support for notebooks
- Experiment tracking and versioning
- Result visualization with Cards
Cloud Scaling
- Effortless horizontal and vertical scaling
- Support for both CPUs and GPUs
- Fast data access for massive parallel workloads
- Gang-scheduled distributed computing
- Reliable failure recovery and checkpointing
Production Deployment
- Easy dependency management
- One-click deployment to production orchestrators
- Support for AWS Step Functions, Argo Workflows, Airflow, and more
- Built-in event triggering for reactive workflows
Get Started
Installation
Install Metaflow in your Python environment
Quickstart
Create your first flow in minutes
Tutorial
Step-by-step guide to Metaflow features
Core Concepts
Understand flows, steps, and decorators
Community & Support
GitHub
Star us on GitHub and contribute
Slack Community
Join our community workspace for help
Documentation
Official Metaflow documentation
Blog
Read about Metaflow at Netflix
From Prototype to Production
Metaflow unifies code, data, and compute at every stage, ensuring seamless end-to-end management of real-world AI and ML systems. Whether you’re working on classical statistics or state-of-the-art deep learning and foundation models, Metaflow adapts to your workflow.Ready to build your first Metaflow project? Head to the Installation guide to get started!