ML Software Engineering

By the time a model matters, it’s living inside a code base – versioned, tested, and shipped by people who weren’t in the room when you wrote the notebook. This chapter is about making that transition smooth: a project structure that scales, a single source of truth in pyproject.toml, guardrails from ruff and mypy, a thin CLI over a clean SDK (with typer), a respectful relationship between notebooks and repos, and environments you can recreate on a fresh laptop without drama.