Landing page
Dashboard
Kanban board layout for lead tracking
Donation Screen
Instant Solana Payment

Donair - Instant agentic crowdfunding

🧐 Inspiration

Event-based fundraising is slow. You need big-ticket sponsors and manual donor profile research. We built Donair to make that process agentic, instantaneous, and versatile. Simply describe your event: Donair identifies high-potential supporters, enables rapid micro- to macro-scale transactions, and builds hype through real-time rewards.

🤯 What it does

Donair is an agentic platform that transforms each fundraising, donation matching, and event planning on any scale.

Before your event

Tell Donair about your event, fundraising goal, and donation incentives
Sit back as Donair identifies key target donors personalized to your mission

During your event

Hit Live
Watch Donair seamlessly sets up a portal for rapid transactions in any currency and amount
Enjoy audience engagement through live incentive trackers

After your event

Assess analytics across events
Do it all over again (since Donair raised tons of money for you!)

💭 How it works:

AI-Powered Donor Discovery Donair uses Google Gemini AI with vector embeddings to semantically match your event description against a database of >110,000 historical publicly-available philanthropic donations sourced from CharityCAN and Indiana University. The system generates embeddings from your event description, then performs Retrieval-Augmented-Generation (RAG) using vector similarity search in Supabase, along with web search for the latest results, to find recipients with aligned missions. From there, it identifies donors who have previously supported similar causes, filtering by capacity (gift size relative to your goal) and geographic relevance.

Intelligent Agent Chat An interactive AI agent powered by Gemini 2.5 Flash acts as your fundraising assistant. It can:

Search your historical donor database using semantic matching
Perform web searches to discover new prospects
Extract donor information from news articles and public sources
Generate AI insights (capacity, affinity, timeliness scores) for each lead
Research specific donors and provide detailed analysis
All three scores are normalized to 0–100 and clamped to ensure valid ranges. They’re combined using weighted averaging: capacity and affinity each contribute 40%, timeliness contributes 20%, reflecting that financial ability and cause alignment matter more than timing. A timeliness score ≥80 adds a +10 bonus to the final relevance score, then the result is normalized to a 0–0.99 scale for database storage. The weighting prioritizes donors who can give and care about the cause, with timeliness as a tie-breaker for immediate opportunities.

Real-Time Payment Processing During live events, Donair integrates with Solana Pay to enable instant donations. The platform:

Generates QR codes for quick mobile donations
Monitors Solana blockchain transactions in real-time via WebSocket connections
Displays live donation feed with instant updates
Supports major currencies with automatic conversion
Tracks offline donations for comprehensive analytics

Lead Management & Analytics A Kanban-style board helps you manage leads through the pipeline (Discovered → Contacted → Negotiating → Confirmed). Each lead includes:

AI-generated relevance scores (capacity, affinity, timeliness)
Contact information extracted from web sources
Ethics flags (GREEN/YELLOW/RED) for risk assessment
Real-time monitoring signals (awards, appointments, business success)
Historical giving patterns and alignment analysis

News Monitoring & Intelligence Automated scanning discovers new opportunities by:

Monitoring news for awards, board appointments, and business milestones
Tracking existing leads for capacity signals (IPOs, promotions)
Identifying philanthropic activity and grant announcements
Providing actionable suggestions for outreach

👷‍♂️ How we built it

Frontend:

Next.js 16 + React 19

TypeScript

Tailwind CSS 4

Supabase Client - Real-time subscriptions for live updates, authentication, and database queries

Google Gemini AI Integration:

gemini-2.0-flash-exp for donor matching and SQL generation
gemini-2.5-flash for agent chat with tool calling (web search, database queries)
text-embedding-004 for semantic vector search

Solana Pay Integration:

@solana/pay for generating payment URLs and QR codes
@solana/web3.js for blockchain connection and transaction monitoring
WebSocket listeners for real-time transaction detection on Solana mainnet

Recharts - Data visualization for analytics dashboards and fundraising progress

React Context API - State management for authentication and notifications

Backend:

Express.js Server - RESTful API endpoints for donor queries and data processing

PostgreSQL (Supabase) - Relational database with:

Vector similarity search using pgvector extension
Row Level Security (RLS) policies for multi-tenant data isolation
Real-time subscriptions via Supabase Realtime
Custom RPC functions for semantic matching (match_recipients)

CSV Import Pipeline - Batch processing of historical donation data with indexing for fast queries

Google Gemini API - AI-powered features:

Embedding generation for semantic search
Natural language understanding for event descriptions
Structured data extraction from web search results
Multi-dimensional scoring (capacity, affinity, timeliness, ethics)

Key Technical Challenges:

Vector Search Optimization - Implemented cosine similarity search with configurable thresholds (0.3) to balance recall and precision. Used Supabase RPC functions to efficiently query embeddings.
Real-Time Transaction Monitoring - Built WebSocket connection to Solana mainnet to listen for on-chain transactions. Handles connection failures gracefully with fallback demo mode.
AI Agent Tool Calling - Designed a flexible tool system where Gemini can call functions like find_similar_recipients, search_web, get_event_leads, and save_lead. Structured prompts ensure the agent only creates leads when explicitly requested.
Multi-Currency Support - Implemented SOL-based conversion using current exchange rates, allowing users to think in dollars with Solana's rapid crypto payments.
Lead Deduplication - Smart matching algorithm prevents duplicate leads by comparing names and checking existing database entries before insertion.

Architecture Highlights:

Serverless API Routes - Next.js API routes handle all backend logic, eliminating need for separate deployment
Real-Time Updates - Supabase Realtime channels provide instant updates for donations, lead status changes, and event status
Progressive Enhancement - Works offline with cached data, gracefully degrades when WebSocket connections fail