# Netflix/metaflow

Build, Manage and Deploy AI/ML Systems

Repository: https://github.com/Netflix/metaflow
Canonical: https://ross.abutalabs.com/products/metaflow
Homepage: https://metaflow.org
Language: Python
License: Apache-2.0
License Family: permissive
Topics: machine-learning, model-management, ai, ml, ml-platform, ml-infrastructure, python, mlops, datascience, high-performance-computing, kubernetes, aws, azure, gcp, llm, llmops, agents, generative-ai, cost-optimization, distributed-training
Last push: 2026-08-25T23:19:42+00:00

## Health v2 (maintenance only)
Score: 95/100 (v2, computed 2026-09-03T02:39:23.370411+00:00)
- activity 99, release rhythm 86, longevity 100
- inputs: {"age_days": 2542, "days_push": 8, "days_rel": 16, "gap_med": 4, "n_releases_24m": 126}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 10245, forks 1336 (observed 2026-08-28T04:10:41.272146+00:00)

## What it is
Metaflow is a human-centric Python framework from Netflix for building, managing, and deploying real-life AI/ML and data science systems. It unifies local prototyping, experiment tracking, scalable cloud compute (CPUs/GPUs on Kubernetes or AWS Batch), dependency management, and one-click production deployment in a single API.

## Use cases
- orchestrate machine learning pipelines in plain python
- scale model training to gpus in the cloud
- track experiments and version artifacts automatically
- deploy ml workflows to production with one command
- run thousands of parallel data processing tasks
- prototype in notebooks then move to production without rewriting
- build and deploy llm training or agent workflows

## When to choose
- your team needs to take ML prototypes from laptop to production reliably
- you want experiment tracking, versioning, and orchestration in one pythonic API
- you need to scale compute across cloud providers or Kubernetes without changing code
- you want a battle-tested framework proven at Netflix-scale

## When to avoid
- you only need simple cron-style scheduling without ML workflow features
- your project is a lightweight ETL script with no scaling or tracking needs
- you require a visual drag-and-drop pipeline builder rather than code-first workflows
- you cannot deploy any infrastructure, since cloud features require a Metaflow stack

## Facets
- artifact type: framework
- maturity: stable
- function: machine-learning, workflow-automation, scheduling, etl, llm-training, data-science, developer-tools
- domain: machine-learning, data-science, large-language-models, cloud-computing
- platform: python, cloud
- tags: mlops, ml-platform, experiment-tracking, workflow-orchestration, llmops, distributed-computing, pipelines, model-management, ai-agents, devops, kubernetes, macos, linux, docker

## Member repositories
- Netflix/metaflow (main) score 95

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:10:41.272146+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:19:24.880074+00:00, confidence not recorded.
  - readme: https://github.com/Netflix/metaflow (fetched 2026-08-28T04:10:41.272146+00:00, sha 389ef18c0171)
  - homepage: https://metaflow.org (fetched 2026-08-29T08:18:57.476344+00:00, sha a39f2b334088)
  - site_page: https://docs.metaflow.org/getting-started/tutorials (fetched 2026-08-29T08:18:57.479143+00:00, sha 4c0568745fce)
  - site_page: https://docs.metaflow.org/ (fetched 2026-08-29T08:18:57.481030+00:00, sha a3a928226a18)
  - site_page: https://docs.metaflow.org/getting-started/install (fetched 2026-08-29T08:18:57.482773+00:00, sha 3c7cb6c595da)
  - site_page: https://docs.metaflow.org/getting-started/devstack (fetched 2026-08-29T08:18:57.484332+00:00, sha 5e658873c7a1)
  - site_page: https://docs.metaflow.org/scaling/remote-tasks/introduction (fetched 2026-08-29T08:18:57.485904+00:00, sha 585e14a2d3dd)
  - registry_pypi: https://pypi.org/pypi/metaflow/json (fetched 2026-08-29T08:18:57.487597+00:00, sha cafcff52840a)
- Data as of 2026-08-30T08:39:29.467469+00:00.