🏆 Built at AGI Ventures Canada Hackathon 3.0
| 20,000+ Lines of Code |
2 😊 Lines Written by Hand |
$400 AI Tokens Used |
12 Hours to Build |
This entire platform was built using Claude Code - we literally wrote only 2 lines of code by hand. Everything else was AI-generated: the complex multi-agent system, React components, API endpoints, CSS styling. This represents the future of software development.
HackRadar is a revolutionary platform that transforms how hackathons evaluate projects. Teams get real-time AI feedback to improve their submissions, while organizers get consistent, fair evaluation across all projects.
- 🎨 Multi-Format Support: Upload PDFs, screenshots, websites, pitch decks
- 🧠 AI-Powered Analysis: Multi-agent system with specialized evaluation agents
- 📊 Real-Time Scoring: Instant feedback across 6 key criteria
- 💡 Actionable Feedback: Specific suggestions on HOW to improve
- 📈 Progress Tracking: See your score evolution over time
- 🏆 Live Leaderboard: Track competition in real-time
A unique collaboration of three minds with one vision:
- Ciprian (Chip) Rarau - DevOps, deployment, AI coding implementation
- Yehor Sanko - AI architect, prompt engineering, UI/UX refinement
- Luca Rarau - Customer validation, sales, product feedback
The magic happened when Ciprian and Yehor discovered they had the same idea independently - creating a tool to help hackathon teams succeed.
🎨 Frontend Layer
├── Next.js 15
├── React 19
├── TypeScript
└── Tailwind CSS
⚡ API Layer
├── /api/projects
├── /api/timeline
├── /api/assess
└── /api/leaderboard
🧠 AI Evaluation Engine
├── BaseAgent (Orchestrator)
├── TextEvaluator (Claude 3.5)
└── SRTracker (Checklist)
💾 Database Layer
├── MongoDB Atlas
├── Projects Collection
├── Timeline Collection
└── Evaluations Collection
Revolutionary cumulative scoring that maintains context across multiple submissions. Teams' scores can only improve or maintain - never decrease unfairly.
// Conversation continuity across evaluations
const messageHistory = await buildMessageHistory(projectId);
const evaluation = await evaluateWithContext(
content,
messageHistory,
conversationId
);Timeline-based architecture enables perfect audit trails and historical tracking.
Innovative score anchoring system ensures teams are protected from scoring volatility.
| Criterion | Points | Description |
|---|---|---|
| Clarity | 15 | Message clarity and structure |
| Problem Value | 20 | Pain point identification |
| Feasibility | 15 | Technical evidence |
| Originality | 15 | Innovation factor |
| Impact | 20 | Conversion potential |
| Readiness | 15 | Completeness check |
- Node.js 18+
- MongoDB connection string
- Anthropic API key
# Clone the repository
git clone https://github.com/crarau/HackRadar.git
cd HackRadar/hackradar-next
# Install dependencies
npm install
# Set up environment variables
cp .env.example .env.local
# Edit .env.local with your credentials
# Run development server
npm run dev
# Open http://localhost:7843MONGODB_URI=your_mongodb_connection_string
ANTHROPIC_API_KEY=your_anthropic_api_key
NEXT_PUBLIC_GOOGLE_CLIENT_ID=your_google_oauth_client_id| Hours | Achievement | Details |
|---|---|---|
| 0-2 | Domain & Infrastructure | Deployed on Azure/Vercel, configured domain |
| 2-4 | Core Development | Next.js app, authentication, database schema |
| 4-6 | AI Multi-Agent System | Claude API integration, evaluation agents |
| 6-8 | Frontend & UX | React components, real-time updates |
| 8-10 | Customer Validation | Talked with 10+ teams, gathered feedback |
| 10-12 | Testing & Polish | End-to-end tests, API simulation, UI testing |
During the hackathon, we:
- 🗣️ Interviewed 10+ teams directly
- 👀 Saw teams' eyes light up when receiving actionable feedback
- 🎯 Pitched to existing teams and watched them improve in real-time
- ✅ Validated the need for real-time feedback during hackathons
We leveraged the power of LLMs to accelerate our testing strategy:
- End-to-End Tests for main application flows
- API Simulation to test without external dependencies
- UI Testing through simulated user interactions
- LLM-Assisted Testing: Used AI to write tests and validate API responses directly
This approach helped us accelerate development significantly. While we don't have 100% coverage, we ensured the critical paths were thoroughly tested.
- Frontend: Next.js 15, React 19, TypeScript, Tailwind CSS
- Backend: Next.js API Routes, Azure VM deployment
- Database: MongoDB Atlas
- AI: Anthropic Claude 3.5 Sonnet
- Authentication: Google OAuth
- Development: Claude Code (99% AI-generated!)
- 60 Git Commits
- 20,000+ Lines of Code
- 2 Lines Written Manually 😊
- $400 Worth of AI Tokens
- 10+ Teams Interviewed
- End-to-End Tests for Main Features
- Get specific, actionable feedback to improve your project
- Track your progress throughout the hackathon
- Understand exactly what judges are looking for
- Iterate quickly based on AI suggestions
- Consistent evaluation across all projects
- Reduce judge workload and bias
- Get comprehensive analytics
- Provide better experience for participants
- Submit Your Project: Upload any format - PDF, screenshots, code, URLs
- Get Instant Analysis: AI evaluates across 6 key criteria
- Receive Actionable Feedback: Specific suggestions for improvement
- Track Progress: See your score evolution over time
- Compete Live: Watch the leaderboard update in real-time
# Production build
npm run build
# Run tests
npm test
# Deploy to Vercel
vercel --prod- 🌐 Live Platform: hackradar.me
- 📖 Technical Journey: hackradar.me/technical-journey
- 🏆 Leaderboard: hackradar.me/public-dashboard
MIT License - see LICENSE file for details
- AGI Ventures Canada for hosting the hackathon
- Anthropic for Claude API
- All the teams who provided invaluable feedback
- The hackathon community for the inspiration
Built with ❤️ and AI by the HackRadar Team
We didn't just build a tool. We built the future of hackathon feedback.
