RD Agent

RD Agent is a self-hosted AI research agent for data scientists, researchers and quantitative developers. It runs experiments on a Linux machine or server, while chat and embedding requests use providers through LiteLLM, including OpenAI and Azure OpenAI. Those remote configurations require API credentials and internet access; hosting the agent yourself doesn't move their inference onto your hardware.

The framework separates proposing ideas from implementing them. It generates runnable code, executes experiments and uses the results to refine subsequent proposals. This makes it suited to repeated model and feature experiments rather than a single conversation about a dataset.

Its quantitative finance workflows use Qlib to develop factors and models together or separately. It can extract factors from financial reports and implement model structures described in research papers. Kaggle workflows cover model tuning and feature engineering.

FT-Agent handles autonomous language-model fine-tuning, including data processing, training, evaluation and feedback-guided refinement against a target benchmark. The project also includes Agent² RL-Bench for evaluating agents that build post-training pipelines with static optimization and online reinforcement learning.

RD Agent is open source under MIT, and most scenarios require Docker. A self-hosted browser interface starts tasks and displays execution traces, with authenticated remote access available. That interface excludes the data science scenario, which uses a separate Streamlit log viewer.