Hack-Nation's Market Link Network
AIPythonFastAPID3.jsOpenAIGemini
Winner 'Creativity Award' & Top 10 Overall. An AI-powered platform uncovering hidden correlations in prediction markets.
Project Overview
Built in 39 hours, this project won the 'Creativity Prize' and placed in the Top 10. It combines OpenAI embeddings, k-NN similarity, and Gemini reasoning to reveal arbitrage opportunities across financial prediction platforms.
Demo Videos
How It Works
- Data Streams → Vector Intelligence: Converting market descriptions into 1536-dimensional vectors using `text-embedding-3-large`.
- Embedding & Clustering: Using HDBSCAN and cosine similarity to group logically related markets.
- Deep Research Layer: Gemini reasoning checks for logical exclusivity, correlations, and arbitrage opportunities.
- Expected-Value Engine: Calculating joint probabilities and expected profits for each combination.
System in Action
Interactive Network Visualization
Force-directed graph showing market relationships. Node size = volume, edge thickness = correlation strength. High-volatility markets pulse to indicate rapid price changes.
Market Correlation Analysis
AI-discovered correlation analysis revealing hidden dependencies.
Expected Value & Returns
Complete EV calculation and risk assessment for identified arbitrage opportunities.
Tech Stack
Backend & AI
- Python FastAPI: High-performance async API
- Supabase: Postgres with vector extensions
- OpenAI Embeddings: Vector generation
- Google Gemini: AI reasoning engine
Frontend
- Next.js 14: React framework
- D3.js: Interactive force-directed graph
- TypeScript: Type safety
- Tailwind CSS: Styling
