autogluon/autogluon-assistant

What it solves

AutoGluon Assistant (MLZero) automates the entire end-to-end process of creating machine learning and deep learning solutions. It removes the need for human intervention to transform raw multimodal data into high-quality ML models.

How it works

It utilizes a multi-agent system that can execute LLM-generated code to build ML workflows. The system supports multiple LLM providers (such as AWS Bedrock, OpenAI, Anthropic, and SageMaker) and can be operated via a CLI, Web UI, or the Model Context Protocol (MCP). For complex tasks, it employs a node-based manager and Monte Carlo Tree Search (MCTS) to refine its approach.

Who it’s for

Data scientists and ML researchers who want to automate the repetitive parts of the ML pipeline or those looking for a zero-intervention approach to building multimodal ML solutions.

Highlights

• End-to-End Automation: Handles everything from raw data to final ML solution.
• Multimodal Support: Capable of processing multimodal data for deep learning workflows.
• Multiple Interfaces: Accessible via CLI, Web UI, and MCP.
• Agentic Search: Uses MCTS and a node-based manager for improved performance in finding the best ML strategy.
• Conversational Guidance: Includes a "Chat Mode" for ML advice without executing code.