AIDE ML (AIDE: AI-Driven Exploration) is an open-source ML engineering agent that uses agentic tree search to iteratively generate, debug, and benchmark code. It ships as a Python package with CLI, visualization, and utilities for reproducible experiments.
Main Features
- Agentic tree-search for automated code drafting and improvement.
- CLI and Streamlit Web UI for running experiments and visualizing solution trees.
- Support for local or remote LLM backends and integrated logging for reproducibility.
Use Cases
- Automating ML experiment construction and optimization.
- Research on agent architectures and search strategies for code generation.
- Rapidly producing reproducible code improvements and baselines in controlled environments.
Technical Highlights
- Implemented in Python, available via pip and Docker for easy deployment.
- Offers tree visualization, log management, and exportable best-solution artifacts.
- MIT license, active community, and accompanying paper (arXiv) and examples.