🧠 Agentic-DS: Agentic System for Data Science Teams

Agentic-DS is a modular, agent-based system that simulates the behavior of a data science team. It uses autonomous agents—each with a specific role like EDA, feature engineering, modeling, or monitoring—to collaboratively tackle machine learning tasks.

🚀 Tested on the Kaggle Playground Series - Season 5 Episode 4
📉 Achieved a Root Mean Squared Error (RMSE) of 13.5, demonstrating promising early performance for an automated agentic setup.

---

📁 Project Structure

Agentic-DS/
├── agents/             # Specialized agents for tasks (EDA, Modeling, etc.)
├── config/             # YAML configuration files for agents and teams
├── data/               # Input datasets (placeholder)
├── models/             # Trained model artifacts
├── outputs/            # Output logs, predictions, and results
├── tools/              # File tools, shell tools used by agents
├── main.py             # Main script to launch the agentic workflow
├── requirements.txt    # Python dependencies
└── README.md           # Project documentation

---

🚀 Getting Started

1️⃣ Clone the Repository

git clone https://github.com/Abhir1902/Agentic-DS.git
cd Agentic-DS

2️⃣ (Optional) Create a Virtual Environment

python -m venv venv
source venv/bin/activate  # On Windows use: venv\Scripts\activate

3️⃣ Install Dependencies

pip install -r requirements.txt

4️⃣ Run the Agentic System

python main.py

This will execute the entire pipeline as configured in config/team.yaml.

---

🧠 What is Agentic-DS?

Agentic-DS is an agent-based orchestration system for automating data science workflows. Each agent:

• Has a dedicated role (EDA, Feature Engineering, Modeling, Monitoring)
• Uses tools like shell commands and file operations
• Can communicate with other agents
• Makes decisions based on defined goals

The system is inspired by how real-world data science teams collaborate on complex problems.

---

📊 Experiment Details

• Dataset: Kaggle Playground Series - S5E4
• Task: Predict podcast popularity score
• Approach: Autonomous agents handling data cleaning, feature engineering, model selection, and evaluation
• Model Used: Gradient Boosting Regressor
• Result: Achieved 13.5 RMSE on the validation set

---

🧰 Tech Stack

• Python 3.10+
• Agno (Agent Framework)
• scikit-learn
• pandas, numpy, matplotlib
• YAML configuration for teams and roles

---

🛣️ Roadmap

• Add logging and dashboards for agent activity
• Integrate with LLMs (e.g., OpenAI, LangChain)
• Support classification/regression/task auto-selection
• GUI for configuring agent teams and workflows
• Benchmark across multiple datasets

---

🤝 Contributing

Contributions welcome!

1. Fork the repo
2. Create your feature branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

---

📬 Contact

• Twitter: @AbhirMirikar
• LinkedIn: Abhir Mirikar
• GitHub: Abhir1902

---

📄 License

This project is licensed under the MIT License. See the LICENSE file for details.

---

⭐️ If you like this project, please consider starring it on GitHub!