At Lyvena, we focus on applied AI: building systems that solve real operating bottlenecks, ship cleanly, and hold up in production. Today, I'm excited to share Prospera Havens, a real estate platform built for the Próspera ZEDE in Roatán, Honduras.
What We Built
A full-stack real estate platform combining AI-powered property matching with fractional ownership tokenization, built for the Próspera ZEDE jurisdiction.
The Challenge
Próspera ZEDE is one of the world's most innovative special economic zones. It attracts forward-thinking residents and investors who expect cutting-edge tools for every aspect of life — including real estate. The existing property market lacked a digital platform that could handle:
- AI-driven property matching — buyers shouldn't have to scroll through thousands of listings. They should be able to describe what they need and get matched instantly.
- Fractional ownership support — the ZEDE model attracts both full-time residents and part-time investors, many of whom want to co-own high-value properties.
- Transparent transaction history — every offer, counter, and closing should be auditable without requiring a title company visit.
The Architecture
We built Prospera Havens on a Next.js front-end with a FastAPI backend, PostgreSQL for structured data, and Weaviate for vector search. The AI layer uses fine-tuned GPT-4 for natural language property queries and CLIP embeddings for visual similarity search — so users can say "find me a beachfront studio with high ceilings" or upload a photo of an interior they like and get relevant matches.
The blockchain layer uses NEAR Protocol for fractional ownership smart contracts. Each property is represented as an NFT that can be fractionally divided into up to 1,000 fungible tokens. This lets investors buy as little as 0.1% of a $2M villa.
Next.js
Frontend framework
FastAPI
Backend API
Weaviate
Vector search
The AI Matching Engine
The heart of Prospera Havens is the AI matching engine. When a user submits a query — either as text or an image — the system:
- Parses intent using GPT-4, extracting key requirements (location, price range, property type, amenities).
- Generates embeddings for both the query and all available listings using OpenAI's text-embedding-3-large model.
- Performs hybrid search — combining vector similarity with structured filters (price, size, bedrooms) in Weaviate.
- Ranks and explains — showing users not just the best matches, but why each property was recommended.
Fractional Ownership with NEAR
For fractional ownership, we deployed smart contracts on NEAR Protocol using Rust. Each property is minted as an NFT, then a fractionalization contract splits ownership into fungible tokens. Smart contracts handle:
- Property tokenization and verification
- Primary offering of fractional shares
- Secondary market trading between verified investors
- Automated dividend distribution for rental income
- Voting mechanisms for major property decisions
Results
Prospera Havens launched in April 2026. In the first two months, the platform listed 47 properties, processed 18 successful fractional ownership transactions, and matched 142 prospective buyers with properties they wouldn't have found through traditional searches. The AI matching engine reduced average search-to-offer time from 14 days to 3.
More importantly, the platform demonstrated that applied AI — not flashy demos, but production systems — can fundamentally reshape how real estate markets operate in special economic zones.
“This isn't about replacing real estate agents. It's about giving them tools that make every interaction smarter, faster, and more transparent.”
The Prospera Havens model is now being explored for expansion to other ZEDEs and special economic zones globally. If you're interested in bringing this approach to your jurisdiction, let's talk.