Getting Started with RD-Agent: A Practical Tutorial for ML Researchers

Getting Started with RD-Agent: A Practical Tutorial

In this tutorial, we'll walk through setting up RD-Agent and running your first automated research workflow. Whether you're an ML researcher or a data scientist, these steps will help you get productive in minutes.

Step 1: Pull and Run the Docker Image

RD-Agent ships as a Docker image, making deployment straightforward:

docker pull jidodata/rd-agent:latest
docker run -d -p 8080:8080 -v ./rd-agent-data:/data jidodata/rd-agent:latest

This starts the RD-Agent web interface on port 8080 with persistent storage in the ./rd-agent-data directory.

Step 2: Configure Your First Experiment

Once the web UI is running, navigate to http://localhost:8080. You'll see the Streamlit dashboard where you can:

• Define a research goal — describe what you want to investigate in natural language
• Select data sources — connect to CSV files, databases, or APIs
• Choose model type — regression, classification, or custom pipelines

Step 3: Run and Monitor

RD-Agent's agent framework handles the rest — it will:

1. Analyze your data and generate preprocessing steps
2. Explore multiple model architectures
3. Optimize hyperparameters automatically
4. Log all results with full metadata for reproducibility

You can monitor progress in real-time through the Streamlit dashboard and compare results across experiments.

Step 4: Review Results

After the agent completes its work, you'll get a comprehensive report including model performance metrics, visualizations, and exportable results. All experiment metadata is stored locally for later analysis.

Use Cases