PyCaret
Low-code AutoML engine and self-hosted machine learning platform with a React web interface and FastAPI backend.
Train models with Scikit-learn, manage experiments in a React UI, and deploy production API endpoints.
PyCaret is an open-source AutoML library and self-hosted machine learning control plane written in Python. It provides a web UI to train algorithms, track experiments, register models, and serve prediction APIs locally or on your servers. It suits data scientists and predictive analysts seeking a unified workspace without commercial SaaS platforms.
What it is
The tool combines a Scikit-learn engine with a FastAPI backend and React dashboard to automate model training across classification, regression, clustering, and time series tasks. It manages experiment runs, tracks model versions in an integrated registry, and provides REST endpoints for live prediction requests. Pipeline state persists locally using SQLite and file storage, with optional support for external PostgreSQL and S3 backends.
Who it's for
This platform suits data scientists, predictive analysts, and machine learning engineers who want to automate experiment tracking and model deployment on their own infrastructure. It is less suitable for teams requiring fully decoupled, large-scale cloud microservices today, as version 4.0's distributed worker architecture is still under development.
Key features
Integrated React Control Plane
Configure experiments, inspect leaderboards, promote models, and test predictions through a dark-mode web dashboard.
Automated Scikit-Learn Engine
Train and evaluate multiple machine learning algorithms concurrently across tabular data tasks.
Self-Hosted Prediction Endpoints
Serve deployed models directly through FastAPI using standard HTTP POST requests with JSON payloads.
LLM Dataset Copilots
Connect Anthropic or OpenAI API keys to receive automated dataset insights and experimental guidance inside the UI.
Pluggable Infrastructure Backends
Swap local SQLite and file system storage for PostgreSQL and Amazon S3 using environment variables.
Up and running in 3 steps
Our honest verdict
What's great
- Combines AutoML training, experiment tracking, and model serving in a single containerized stack
- Point-and-click UI allows complete workflow configuration without writing YAML files
- Provides immediate curl-compatible prediction endpoints once models are promoted
- Runs locally on a laptop using a single Docker Compose command
Where it falls short
- Version 4.0 is a work-in-progress refactor that introduces breaking changes over 3.x
- Requires substantial local RAM resources (6GB+ recommended) during Docker container builds
- Advanced distributed cloud scaling currently requires manual configuration of underlying protocol interfaces
- Database encryption keys and stored credentials cannot be recovered if the local volume is reset
Pricing & license
PyCaret is free open-source software available via GitHub. There are no paid commercial tiers offered by the maintainers; your primary expense is the infrastructure cost of running the server containers.
Open remote AI-training roles
874 open remote AI-training roles, updated November 15, 2026; most pay $20–$170 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
What database does PyCaret 4.0 use out of the box?
It uses an in-process SQLite database stored inside a Docker volume, which can be configured to use PostgreSQL for production.
Can I run PyCaret locally without Docker?
Yes, you can install dependencies using the uv package manager and run the FastAPI backend and React frontend as separate processes.
