FairSplit AI

FairSplit AI is an AI-powered expense splitting platform that eliminates the awkwardness of money between friends.

Comment

Below is a clean, hackathon-ready write-up for the requested sections, fully synthesized from your FairSplit AI content and phrased in a concise, judge-friendly style.

💡 Inspiration

FairSplit AI was inspired by a universal, painfully familiar problem: money awkwardness between friends. From group dinners to shared trips, small unpaid balances often lead to discomfort, resentment, or one person quietly overpaying. Existing apps track debts, but they don’t solve the human side of the problem. We wanted to build something that not only splits bills correctly, but also protects friendships by making settlements fair, efficient, and emotionally intelligent.

⚙️ What It Does

FairSplit AI is an AI-powered group expense splitting platform that:

  • Scans receipts automatically using Vision AI
  • Calculates who owes what with zero manual input
  • Optimizes settlements to reduce many transactions into just a few
  • Analyzes group fairness and spending imbalance
  • Provides emotionally intelligent suggestions to prevent resentment

In short, it turns messy group expenses into clear, fair, and friendship-safe outcomes.

🛠️ How We Built It

We built FairSplit AI as a production-ready, full-stack application with a multi-agent AI architecture.

Core components:

  • ReceiptScannerAgent – Uses GPT-4 Vision or local Ollama models to extract receipt data
  • DebtOptimizerAgent – Applies graph algorithms to minimize payment transactions
  • EmotionalIntelligenceAgent – Analyzes fairness, detects imbalance, and suggests corrective actions

Tech Stack:

  • Frontend: Next.js 14, React 18, TypeScript, Tailwind CSS
  • Backend: Next.js API Routes, Prisma ORM
  • Database: PostgreSQL
  • AI: OpenAI GPT-4o or free local Ollama models
  • Cache: Redis with semantic caching
  • Auth & Security: NextAuth, PII detection, moderation, rate limiting

We focused heavily on clean architecture, test coverage, security guardrails, and real-world deployability.

🚧 Challenges We Ran Into

  • Receipt variability: Real receipts are messy, inconsistent, and poorly formatted
  • Debt minimization logic: Designing an algorithm that was both mathematically optimal and easy to explain visually
  • AI cost control: Solved using semantic caching and optional local AI models
  • Fairness modeling: Translating emotional imbalance into measurable signals without being intrusive
  • Hackathon time pressure: Balancing polish, innovation, and stability under tight deadlines

Each challenge pushed us to design smarter, not just faster.

🏆 Accomplishments We’re Proud Of

  • Built a never-before-seen emotional debt optimization system
  • Reduced transaction counts by up to 70%
  • Delivered a live, demoable AI experience in under 5 minutes
  • Implemented 33+ AI design patterns
  • Achieved production-grade security and architecture
  • Enabled 100% free AI usage via local Ollama integration
  • Created a solution judges can immediately relate to and use

This is not a prototype — it’s a real, scalable product.

📚 What We Learned

  • AI is most powerful when combined with human-centered design
  • Optimization problems become far more compelling when visualized
  • Cost-aware AI architecture (caching, local models) is critical for scale
  • Emotional intelligence can be modeled, measured, and made useful
  • Hackathons reward clarity of story as much as technical depth

Most importantly, we learned that small problems, solved deeply, create massive impact.

🚀 What’s Next for FairSplit AI

  • Payment integrations (Venmo, Paystack, Stripe)
  • Mobile app launch
  • Smarter long-term fairness tracking
  • Group behavior insights and analytics
  • API partnerships with payment and expense platforms
  • Public launch and early user acquisition
  • Seed funding and continued product iteration

Our vision is to make FairSplit AI the default intelligence layer behind group payments worldwide.

FairSplit AI isn’t just about splitting bills — it’s about preserving relationships. Because nobody should lose a friend over $5. 💚

Built With

Share this project:

Updates