agentic-ml

An agentic framework that tackles a data scientist's machine learning problems with LLM agents. Instead of a fixed AutoML search space, an agent writes and runs real training code in an experimentation loop, keeps an auditable record of every trial, and exports a lightweight, production-ready model.

Status: early prototype. The first (and currently only) supported task is tabular regression.

Important: due to a name collision on PyPI, this package is published as jcaste05-agentic-ml (not agentic-ml). The import path is unaffected — it's still import agentic_ml. See the Quick example for install commands.

• Quick example — a runnable end-to-end notebook and installation guide.
• Architecture — how the source code is structured and how the research loop interacts with the production model.
• Reference — Library documentation generated from the source code's docstrings.
• GitHub repository.