Building the Trust Layer for the AI Era - Capturing Real Signals Beyond the Hype
a42z Judge is the world's first autonomous, insightful, and fair AI judging system for hackathons built by the team a42z. We're creating a comprehensive evaluation engine that combines multimodal AI analysis with expert knowledge to deliver transparent, explainable, and bias-reduced assessment results. We were honored to win 🏆 runner-up in the AI Agent Track (sponsored by PPIO Cloud) at AdventureX, China’s largest-ever hackathon, selected from over 8,000 applicants.
a42z Judge - Agentic System Architecture
In the AI era, we need a new paradigm for evaluating innovation. Traditional hackathon judging relies heavily on subjective human judgment, leading to inconsistencies and missed opportunities. a42z Judge addresses this by creating a sophisticated AI-powered evaluation system that:
- Cross-references project materials (code, demo decks, PRDs) against open-source repositories and past winning cases
- Leverages quality signals like GitHub stars, activity metrics, and code quality indicators
- Emulates the judgment of tech industry leaders through AI twins
- Delivers structured, multi-dimensional assessments that go beyond simple scoring
- Input Processing: Handles PDF pitch decks and GitHub repository URLs
- Cross-Reference Analysis: Compares against open-source projects and historical data
- Quality Signal Extraction: Analyzes GitHub metrics, code quality, and community engagement
- Real-time Processing: Streams analysis results through webhooks
Business Potential Researcher
- Sources: Crunchbase, TechCrunch, Y Combinator
- Metrics: Fundraising rounds, market share, MRR/ARR, ROI
- Analysis: Idea space capacity, profit margins, market trends
Code Quality Researcher
- Sources: GitHub API, RepoIntel, SonarQube
- Metrics: System robustness, trending/stars, commits/issues
- Analysis: Code metadata, paradigms, documentation quality
Technical Homeomorphism Researcher
- Sources: Dockerfiles, package.json, requirements, configs
- Metrics: Novelty of core technology, cross-domain applicability
- Analysis: Tech sophistication, problem-solving impact, scalability
AI twins of tech industry leaders providing structured evaluations:
- Andrew Ng (Deeplearning.ai) - AI/ML expertise
- Paul Graham (Y Combinator) - Startup evaluation
- Feifei Li (ImageNet) - Computer vision and research
- Sam Altman (OpenAI) - AI innovation and scaling
- Scrapers: X/Twitter, Blogs, YouTube
- Processing: Trunking, Embedding, Reranking
- Storage: Vector Database for semantic search
- Live API Calls: Simulated external service integrations
- Progress Tracking: Real-time workflow step updates
- Webhook Integration: Dify API for advanced AI analysis
- Database Storage: Supabase for persistent results
- Magic UI Components: Animated cards, buttons, and effects
- Terminal Animations: Realistic API call simulations
- Responsive Design: Works across all devices
- Dark Theme: Modern, professional interface
- Multi-dimensional Scoring: Technical, business, innovation metrics
- Structured Comments: Bilingual (Chinese/English) expert feedback
- Debate Simulation: AI twins discuss and reach consensus
- Transparent Process: Full visibility into evaluation criteria
- Next.js 15.2.4 - React framework with App Router
- TypeScript 5.0 - Type-safe development
- Tailwind CSS 4 - Utility-first styling
- Framer Motion - Smooth animations and transitions
- Dify API - AI workflow orchestration
- Supabase - Database and authentication
- GitHub API - Repository analysis
- Webhooks - Real-time data flow
- Magic UI - Custom animated components
- Radix UI - Accessible component primitives
- Lucide Icons - Beautiful iconography
- Shiki - Syntax highlighting
- Crunchbase API - Business intelligence
- TechCrunch API - News and trends
- RepoIntel API - Repository intelligence
- SonarQube API - Code quality analysis
a42z Judge - Video Demo
- Node.js 18+
- npm, yarn, or pnpm
- Supabase account
- Dify API access
- Clone the repository
git clone https://github.com/your-org/a42z-judge.git
cd a42z-judge- Install dependencies
npm install
# or
yarn install
# or
pnpm install- Configure environment variables
cp .env.example .env.localEdit .env.local with your credentials:
# Supabase Configuration
NEXT_PUBLIC_SUPABASE_URL=your_supabase_project_url
NEXT_PUBLIC_SUPABASE_ANON_KEY=your_supabase_anon_key
SUPABASE_SERVICE_ROLE_KEY=your_supabase_service_role_key
# Dify Configuration
NEXT_PUBLIC_DIFY_API_URL=https://api.dify.ai/v1
NEXT_PUBLIC_DIFY_API_KEY=your_dify_api_key
NEXT_PUBLIC_WEBHOOK_URL=https://your-domain.com/api/webhook/dify- Set up the database
-- Run the SQL from SUPABASE_SETUP.md
-- This creates the judge_comments table and necessary indexes- Start the development server
npm run dev
# or
yarn dev
# or
pnpm dev- Open your browser Navigate to http://localhost:3000
- Access your Dify workflow at:
https://cloud.dify.ai/app/your-app-id/develop - Configure HTTP Request tool with webhook URL:
https://your-domain.com/api/webhook/dify - Set up the analysis workflow for GitHub repository evaluation
- Create a new Supabase project
- Run the SQL commands from
SUPABASE_SETUP.md - Configure Row Level Security (RLS) policies
- Set up API keys and CORS settings
- Deploy to Vercel or your preferred platform
- Configure environment variables
- Set up custom domain (optional)
- Enable webhook endpoints
- Upload Project Materials: Submit GitHub repositories and pitch decks
- Real-time Analysis: Watch as AI researchers evaluate projects
- Expert Feedback: Receive structured comments from AI twins
- Transparent Results: View detailed scoring and reasoning
- Submit Projects: Upload GitHub links and documentation
- Track Progress: Monitor analysis in real-time
- Receive Feedback: Get expert evaluation and suggestions
- Understand Scores: See detailed breakdown of assessments
// Example: Submit a GitHub repository for analysis
const response = await fetch('/api/analyze', {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify({
repo_url: 'https://github.com/user/project',
user_email: '[email protected]'
})
});- Consistent Evaluation: AI-powered analysis reduces human bias
- Scalable Process: Handle hundreds of projects efficiently
- Transparent Results: Clear reasoning for all scores
- Quality Assurance: Cross-reference against industry standards
- Fair Assessment: Objective evaluation criteria
- Detailed Feedback: Expert-level comments and suggestions
- Learning Opportunity: Understand what makes projects successful
- Recognition: Merit-based scoring and ranking
- Innovation Tracking: Identify emerging trends and technologies
- Quality Standards: Establish benchmarks for hackathon projects
- Knowledge Sharing: Learn from successful project patterns
- Ecosystem Growth: Foster better hackathon experiences
- Basic AI judging engine
- GitHub repository analysis
- Expert AI twins
- Real-time UI
- Multi-language support
- Advanced code analysis
- Video demo evaluation
- Team collaboration features
- Hackathon platform integrations
- API marketplace
- Custom evaluation models
- Enterprise features
- Self-improving algorithms
- Domain-specific models
- Predictive analytics
- Automated insights
We welcome contributions from the community! Here's how you can help:
- AI Models: Improve evaluation algorithms
- UI/UX: Enhance user experience
- API Integrations: Add new data sources
- Documentation: Improve guides and examples
- Testing: Add tests and improve coverage
- Follow TypeScript best practices
- Use Prettier for code formatting
- Write meaningful commit messages
- Add tests for new features
- Tech Leaders: Inspired by the insights of Andrew Ng, Paul Graham, Feifei Li, and Sam Altman
- Open Source Community: Built on the shoulders of amazing open-source projects
- Hackathon Community: For feedback and inspiration
- AI Research Community: For advancing the state of AI evaluation
Built with ❤️ for the hackathon community
Empowering innovation through intelligent evaluation