Mlflow

This is your go-to for tracking ML experiments and managing model lifecycles without getting locked into a specific framework. It handles the boring parts: logging parameters and metrics during training, versioning models in a registry, and deploying them to production with stage transitions (Staging, Production, Archived). The autologging feature is legitimately useful since it captures everything from scikit-learn, PyTorch, or TensorFlow training runs without manual instrumentation. Works well for solo projects and teams alike. The UI at localhost:5000 gives you a clean comparison view across experiment runs. Used by 20,000+ organizations, so the workflow patterns are pretty well established at this point.

Install to Claude Code

npx -y skills add orchestra-research/ai-research-skills --skill mlflow --agent claude-code