ShieldScore

Inspiration

ShieldScore was inspired by a simple question: why do users need to expose their most sensitive financial data just to prove they’re trustworthy? With AI everywhere and privacy disappearing, we wanted to show that powerful financial insights and strong user privacy can coexist by the power of zero‑knowledge proofs and Midnight’s confidential smart contract platform.

What it does

ShieldScore provides private, AI‑powered credit and risk assessments without exposing any personal data. All scoring runs locally on the user’s device, and only a zero‑knowledge proof of the result is sent to the Midnight blockchain. Lenders get verified eligibility signals. Users keep full control of their financial information.

How we built it

Built the frontend with Next.js + TypeScript

Designed a lightweight on‑device AI scoring model for credit/risk evaluation

Generated zero‑knowledge proofs using Midnight’s proof server

Created a Compact smart contract to verify proofs on‑chain

Integrated the Lace wallet for secure interactions

Developed a clean, Mastercard‑inspired UI for a smooth user experience

Challenges we ran into

Learning Midnight’s Compact language and proof server workflow under time pressure

Ensuring the AI model stayed lightweight enough to run fully client‑side

Designing a ZK flow that was simple for users but still cryptographically sound

Debugging WSL2, Docker, and SDK setup issues

Balancing privacy, accuracy, and performance in a 48‑hour build

Accomplishments that we're proud of

We’re incredibly proud that we built a fully functional, end‑to‑end private credit‑scoring demo in under 48 hours. ShieldScore successfully performs local AI inference and generates zero‑knowledge proofs that verify creditworthiness without exposing any financial data. We also delivered a polished, production‑style UI that makes a complex cryptographic flow feel simple and intuitive. Beyond the product itself, we proved that AI, DeFi, and privacy can work together seamlessly on the Midnight blockchain. Learning an entirely new ecosystem, mastering Compact, and shipping a complete solution in such a short timeframe is an accomplishment we’re genuinely excited about.

What we learned

This project taught us how to build truly confidential dApps using Midnight and its Compact smart contract language. We learned how to run AI inference securely on‑device and how to design user‑friendly flows around cryptographic concepts that are normally intimidating. We also gained a deeper appreciation for privacy‑first architecture in financial applications, where protecting user data is just as important as delivering accurate results. Finally, we learned how to adapt quickly to new tools, SDKs, and constraints.

What's next for ShieldScore

Next, we plan to expand the AI model to support a wider range of financial scenarios and risk profiles. We want to introduce lender dashboards that allow institutions to verify zero‑knowledge proofs and make decisions without ever accessing user data. We’re also exploring multi‑wallet and multi‑chain support to broaden ShieldScore’s reach across the decentralized ecosystem. Improving proof generation speed and overall UX is another priority, along with exploring real‑world partnerships where privacy‑preserving credit checks could have immediate impact.

Built With

• compact
• css
• docker
• lace
• midnight
• next.js
• node.js
• proof
• react
• server
• typescript
• wallet
• wsl2