Verified project record

microsoft/RD-Agent

RD-Agent is an agent framework from Microsoft that automates data-driven R&D by having AI propose new ideas and implement models, features, and factors as runnable code that evolves through feedback. It covers scenarios such as quantitative strategies, Kaggle competitions, and LLM fine-tuning.

Project overview

RD-Agent leads as the top-performing machine learning engineering agent on MLE-bench and is documented as the first data-centric, multi-agent framework for full-stack quantitative strategy R&D via factor-model co-optimization.

Project type: AI Agent · AI Coding · Model Development
Use cases: Automation
Deployment: Refer to project documentation
License: MIT

Best for

AI engineers, researchers, and data teams on Linux who want an agent framework to automate iterative, data-driven R&D and already have access to configured chat and embedding model APIs.

Key capabilities

Automates the full-stack research and development of quantitative strategies via coordinated factor-model co-optimization.
Autonomous LLM fine-tuning for benchmark-driven domain adaptation, invoked via `rdagent llm_finetune`.

Limitations and risks

RD-Agent currently only supports Linux; it is not usable as-is on Windows or macOS.
Docker must be installed (and usable without sudo) for most scenarios.
The system depends on an external LLM provider API (OpenAI, Azure OpenAI, or DeepSeek) for chat and embedding; local orchestration does not make it offline, and there is no local-only data boundary.

Getting started

Create and activate a conda environment: `conda create -n rdagent python=3.10 && conda activate rdagent`, then `pip install rdagent`. Install Docker (usable without sudo) and verify with `docker run hello-world`. Write a `.env` file with CHAT_MODEL, EMBEDDING_MODEL, and API keys — the demos require a chat model supporting ChatCompletion and json_mode plus an embedding model (e.g. gpt-4o, deepseek-chat, text-embedding-3-small, or BAAI/bge-m3). Run `rdagent health_check` to verify, then start with `rdagent fin_quant`.

Evidence and sources

GitHub project description: We are committed to automating these high-value generic R&D processes through R&D-Agent, which lets AI drive data-driven AI.
README: R&D-Agent aims to automate the most critical and valuable aspects of the industrial R&D process, and we begin with focusing on the data-driven scenarios to streamline the developm…
README: You can directly install the R&D-Agent package from PyPI: ```sh pip install rdagent
README: rdagent fin_quant
README: the first data-centric, multi-agent framework designed to automate the full-stack research and development of quantitative strategies via coordinated factor-model co-optimization.

Source updated at: 2026-09-23T18:08:12Z

Open original GitHub project