MLflow

MLflow — Technical Overview

MLflow is a 100% open-source AI engineering platform (Apache 2.0) designed for ML/AI engineers and data science teams building, evaluating, and deploying LLM applications, agents, and ML models.

Key capabilities for developers:

• LLM & Agent Observability: Production-grade tracing and observability built on OpenTelemetry, giving full visibility into LLM calls, agent chains, and tool invocations.
• Systematic Evaluation: 50+ built-in metrics and LLM-as-judge evaluators for rigorous, reproducible model and prompt assessment.
• Prompt Engineering: Versioning, A/B testing, and optimization workflows for prompt management across environments.
• Unified AI Gateway: A single abstraction layer routing requests across multiple LLM providers, reducing vendor lock-in.
• Experiment Tracking & Model Registry: End-to-end ML lifecycle management — log parameters, metrics, artifacts, and register production-ready models.
• Integrations: Native support for LangChain, OpenAI, PyTorch, OpenTelemetry, and 100+ ecosystem tools.

Deployment & access:

• Fully self-hostable; no managed dependency required.
• API available for programmatic access and automation.
• 28,258 GitHub stars | 30M+ monthly package downloads.
• Founded: 2018.