MLE-agent

Why we included this project

Machine learning engineers burn a surprising share of the week on scaffolding: baseline code, data pipelines, the same setup issues re-debugged on every project. MLE-Agent is an LLM agent that takes over that work and runs an experiment end to end rather than just answering questions in a chat pane. It pulls relevant papers and reference implementations from arXiv and Papers with Code to guide its approach, then runs a debugger-coder loop so the code it writes actually works. The tool supports OpenAI, Anthropic, Gemini, and local Ollama models, and its interactive CLI and weekly report generator fit naturally into a research routine. For a researcher or small team that wants an automated collaborator to build baselines and clear repetitive ML tasks, this is a practical starting point.