A Multi-Agentic AI Framework for Autonomous and Collaborative Data Science Workflows

In the evolving landscape of data science, traditional monolithic automation tools often lack flexibility, interpretability, and adaptability in complex, real-world workflows. This paper proposes a novel multiagentic framework that leverages Artificial Intelligencedriven specialized agents to autonomously and collaboratively manage the entire data science pipelinefrom preprocessing and exploratory data analysis to model selection and performance evaluation. Each agent is designed with domain-specific intelligence and operates both independently and cooperatively within a decentralized architecture, orchestrated via a lightweight communication protocol. The framework promotes modularity, scalability, and human-AI collaboration, while significantly reducing manual intervention and operational overhead. Experiments conducted on real-world datasets demonstrate the system's ability to deliver high-performing models with improved efficiency and transparency. This approach not only redefines AutoML by introducing agent-based specialization but also sets the stage for the next generation of AI-assisted data science platforms.