All work
Case StudyWeb3 · FinTech

ConstructKoin — AI-Underwritten Real-Estate DeFi

Tokenized real-estate lending & underwriting protocol.

2025United KingdomShipped
ConstructKoin — AI-Underwritten Real-Estate DeFi — product screenshot
Core DeliverableTokenized real-estate lending & underwriting protocol
Primary Tech StackSolidity · Hardhat · Ethers.js · Next.js · Python
Project ScopeSmart contracts + dApp + AI underwriting pipeline
Client GeographyUnited Kingdom
01The Narrative

ConstructKoin is a decentralized finance protocol that brings real-estate-backed lending on-chain. Property is tokenized and fractionally funded; loan decisions that used to take weeks are supported by an AI underwriting layer and settled through audited smart contracts. We re-architected the frontend and backend, wired secure APIs across chains, payments and the AI model, and added a centralized CRM so the off-chain business could operate against on-chain truth.

02System Architecture

Solidity contracts, developed and tested under Hardhat, hold the tokenized asset and funding logic. A Next.js dApp uses Ethers.js to orchestrate wallet connections and contract calls across multiple chains, with a connection-pool abstraction so users move between networks without losing state. A Python service runs the AI lending-compliance model and streams risk metrics into the app; a data layer reconciles on-chain events with the off-chain CRM.

03The Technical Challenge

Money on-chain is unforgiving: a reentrancy bug or a mis-priced asset is irreversible. The challenge was combining high-frequency AI risk scoring and multichain UX with the correctness guarantees a lending protocol demands — while keeping approvals fast enough to feel like modern fintech, not a blockchain science project.

04Engineered Solutions
  • Audited smart-contract workflows for fractionalized funding with checks-effects-interactions and reentrancy guards throughout the money path.
  • A multichain wallet connection pool that abstracts network switching and RPC failover, so the dApp stays responsive across chains.
  • A high-frequency data layer that ingests AI compliance metrics and on-chain events into a single reconciled view for underwriting.
  • Secure API bridges between the chain, the payment rails and the Python model, with idempotent event handling so nothing double-executes.

Have a system this demanding?

Start a conversation