# facebook/Ax

Adaptive Experimentation Platform

Repository: https://github.com/facebook/Ax
Canonical: https://ross.abutalabs.com/products/facebook-ax
Homepage: https://ax.dev
Language: Python
License: MIT
License Family: permissive
Last push: 2026-08-19T15:37:07+00:00

## Health v2 (maintenance only)
Score: 95/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 98, release rhythm 87, longevity 100
- inputs: {"age_days": 2762, "days_push": 14, "days_rel": 85, "gap_med": 29, "n_releases_24m": 14}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2792, forks 378 (observed 2026-08-28T04:07:22.179202+00:00)

## What it is
Ax is a Python platform for adaptive experimentation that uses machine learning (Bayesian and bandit optimization via BoTorch) to efficiently explore parameter spaces and find optimal configurations. It manages, automates, and deploys experiments with support for complex search spaces, multiple objectives, constraints, noisy observations, and production-grade storage.

## Use cases
- tune hyperparameters for machine learning models
- run A/B tests with multi-armed bandit allocation
- optimize expensive simulations with few evaluations
- find optimal compiler flags or infrastructure magic numbers
- multi-objective optimization with parameter and outcome constraints
- tune engineering design parameters
- run parallel and asynchronous optimization trials

## When to choose
- evaluations are expensive or limited so random/grid search is wasteful
- you need Bayesian optimization with state-of-the-art algorithms out of the box
- you need multi-objective, constrained, or noisy optimization
- you want production-ready experiment management with MySQL storage and automation

## When to avoid
- your objective function is cheap to evaluate and simple grid or random search suffices
- you need a lightweight single-purpose hyperparameter tuner like Optuna
- you are not working in Python or cannot take on PyTorch/BoTorch dependencies
- you need real-time streaming optimization rather than iterative trials

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, benchmarking, data-science, sdk
- domain: machine-learning, artificial-intelligence, data-science, developer-tools
- platform: python, windows, cross-platform
- tags: bayesian-optimization, hyperparameter-tuning, ab-testing, bandit-optimization, adaptive-experimentation, experimental-design, botorch, multi-objective-optimization, algorithms, linux, macos

## Member repositories
- facebook/Ax (main) score 95

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:22.179202+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 2026-08-30T08:15:15.009111+00:00, confidence not recorded.
  - readme: https://github.com/facebook/Ax (fetched 2026-08-28T04:07:22.179202+00:00, sha fc9da1531d4c)
  - homepage: https://ax.dev (fetched 2026-08-29T09:55:17.009614+00:00, sha 7494b2771b65)
  - site_page: https://ax.dev/docs/tutorials/quickstart (fetched 2026-08-29T09:55:17.018566+00:00, sha a23b2b6a877f)
  - site_page: https://ax.dev/docs/why-ax (fetched 2026-08-29T09:55:17.020641+00:00, sha 558a6b64ee2a)
  - site_page: https://ax.dev/docs/next/why-ax (fetched 2026-08-29T09:55:17.022353+00:00, sha 0c27b5f3225d)
  - site_page: https://ax.dev/docs/1.2.4/why-ax (fetched 2026-08-29T09:55:17.023936+00:00, sha 2249ce22a11c)
  - site_page: https://ax.dev/docs/1.1.2/why-ax (fetched 2026-08-29T09:55:17.025463+00:00, sha 6b50d899f026)
  - site_page: https://ax.dev/docs/1.0.0/why-ax (fetched 2026-08-29T09:55:17.026936+00:00, sha 0e77326adb28)
  - site_page: https://ax.dev/docs/0.5.0/why-ax (fetched 2026-08-29T09:55:17.028338+00:00, sha df0a70db138d)
  - site_page: https://ax.dev/docs/installation (fetched 2026-08-29T09:55:17.029792+00:00, sha 79380dd03918)
- Data as of 2026-08-30T08:39:29.467469+00:00.