MLflow provides four core components for the ML lifecycle: Tracking (logging parameters, metrics, and artifacts from experiments), Models (packaging ML models in a standard format), Model Registry (centralized model store with versioning and staging), and Projects (packaging ML code for reproducible runs).
The platform is framework-agnostic, supporting TensorFlow, PyTorch, scikit-learn, XGBoost, Hugging Face Transformers, LangChain, OpenAI, and virtually any Python ML library. MLflow also includes LLM evaluation tools and a deployments server for serving models via REST API.
MLflow is free and open source under the Apache 2.0 license. Databricks offers a managed MLflow experience integrated with their data lakehouse platform. Self-hosted deployment is straightforward with pip install and supports PostgreSQL, MySQL, or SQLite backends.