pycaret/pycaret

What it solves

PyCaret provides a complete, self-hosted machine learning platform that simplifies the entire ML lifecycle—from data upload and experiment design to model registration and deployment—within a single integrated environment. It eliminates the need to manually manage separate tools for AutoML, experiment tracking, and model serving.

How it works

The platform consists of three main components:

• Engine: An AutoML library based on scikit-learn that handles the actual model training and selection.
• Control Plane: A FastAPI backend that manages workspaces, projects, model registries, deployments, and monitoring.
• Web UI: A React-based interface that allows users to configure experiments and manage models via point-and-click instead of YAML files.

It can be deployed locally using Docker Compose, running a compact setup with a FastAPI process and a React bundle, using SQLite for local data persistence by default.

Who it’s for

It is designed for data scientists and ML engineers who want a professional-grade ML platform they can run on a laptop or a small-team production server without complex infrastructure setup.

Highlights

• End-to-End Workflow: Supports the full loop of Train $\rightarrow$ Register $\rightarrow$ Deploy $\rightarrow$ Predict.
• AutoML Capabilities: Automatically trains and ranks multiple algorithms for classification, regression, clustering, anomaly detection, and time series tasks.
• AI Copilots: Integrates LLM-powered assistants for dataset consultation and experiment design.
• Flexible Architecture: Designed to scale from a single-binary local setup to a distributed cloud architecture using protocols for storage, DB, and compute.