Kaggle Mentor AI – AI-Powered Multi-Agent Mentor for Kaggle Competitions | Google ADK & MCP

Kaggle Mentor AI is an AI-powered multi-agent mentor system designed for Kaggle competitions, built as a submission for the Google × Kaggle AI Agents Intensive Vibe Coding Capstone 2026. The project leverages Google's AI Development Kit (ADK) and Model Context Protocol (MCP) to create an intelligent coaching system that guides competitors through data science challenges. It demonstrates advanced multi-agent orchestration and real-time mentoring capabilities for competitive machine learning.

Key Points

• •Multi-agent architecture using Google ADK for coordinated AI mentoring across different competition domains
• •Integration with Model Context Protocol (MCP) for standardized agent communication and tool access
• •Real-time guidance system that adapts to competitor skill levels and competition stages
• •Leverages Kaggle competition data and historical solutions for contextual mentoring
• •Demonstrates practical application of AI agents in competitive machine learning education
• •Built for Google × Kaggle AI Agents Intensive Vibe Coding Capstone 2026 competition
• •Combines multiple specialized agents for different aspects (EDA, feature engineering, model selection, etc.)
• •Provides actionable feedback and optimization suggestions during competition participation

Workflow Diagram

Concepts