RD-Agent

MicrosoftResearch agents & deep researchListed

About this tool

Microsoft's open-source framework for data-driven R&D automation. A research agent proposes hypotheses, a development agent writes and runs code, and experimental feedback guides iterations. Scenarios include data-science competitions, model tuning, implementing architectures from papers, joint quantitative-factor and model optimization, and model fine-tuning. The README reports MLE-bench results.

  • Feature 1: Research and development loops propose, implement and evaluate hypotheses.
  • Feature 2: Feature engineering and model tuning for data science.
  • Feature 3: Research Copilot extracts and implements models from papers.
  • Feature 4: Joint factor and model optimization for quantitative research.
  • Feature 5: Web and Streamlit interfaces show progress.

Resource information

Maintainer
Microsoft; official company project under the microsoft GitHub organization.
Type
Standalone app
Category
Research agents & deep research
Runs on
Agent application
Who it's for
Data scientists, ML engineers and quantitative researchers automating modeling experiments, paper implementations and factor research.
License at time of listing
MIT
Source
GitHub