Inspiration
Healthcare data is one of the most sensitive types of personal information, yet most AI health platforms require users to upload symptoms, medical history, and vitals to centralized servers. We were inspired by a simple but powerful question: “Can AI healthcare exist without ever seeing private patient data?”
PrivaMed AI was built to solve the growing trust crisis in digital healthcare by combining AI with zero-knowledge cryptography. Our goal was to create a future where users can receive intelligent health guidance while maintaining complete ownership and privacy of their medical data.
What it does
PrivaMed AI is a privacy-first AI healthcare intelligence platform that performs secure medical risk assessment using Zero-Knowledge Proofs (ZKPs).
Users enter symptoms, vitals, and health conditions locally on their device. A Compact ZK circuit processes the information privately and generates only a risk classification such as LOW, MEDIUM, or HIGH — without exposing any raw medical data.
The AI model receives only the risk level and provides health recommendations without ever accessing personal health information. This creates a completely privacy-preserving AI healthcare workflow.
Key features include:
Zero-knowledge health risk assessment Local symptom and vitals processing AI-powered health recommendations Privacy-preserving ledger updates GDPR-friendly architecture No storage of personal medical data How we built it
We built PrivaMed AI using a full-stack privacy-focused architecture powered by Midnight Network technologies.
Tech Stack Compact Language for writing ZK smart contracts PLONK proving system over BLS12-381 for cryptographic proof generation TypeScript + JavaScript for frontend logic Node.js backend for secure AI proxy handling Groq API for AI-generated medical recommendations Vanilla HTML/CSS/JS frontend for lightweight deployment Workflow User inputs symptoms and vitals locally Data is converted into Compact witness values A ZK circuit computes the health risk privately Only the proof and risk level are disclosed AI generates recommendations using only the risk category
This architecture ensures that no sensitive medical information ever leaves the user’s device.
Challenges we ran into
One of the biggest challenges was designing a system where AI could still provide meaningful recommendations without direct access to raw health data.
Other major challenges included:
Learning and implementing Zero-Knowledge Proof workflows Designing Compact circuits for medical risk classification Managing secure communication between frontend, proof system, and AI backend Balancing privacy with useful AI outputs Simulating real-world healthcare logic within ZK constraints Ensuring the frontend remained lightweight and easy to demo during hackathon conditions
We also had to carefully architect the system so that absolutely no personally identifiable medical information could accidentally leak into logs, APIs, or blockchain storage.
Accomplishments that we're proud of
We are proud that we successfully built an end-to-end working prototype of a privacy-preserving AI healthcare platform.
Some highlights:
Built a fully functional ZK-powered healthcare workflow Created a novel “ZK-gated AI” architecture Achieved local-only sensitive data processing Integrated AI recommendations without exposing patient information Designed a polished and interactive frontend demo Built a strong business case around privacy-first healthcare AI
Most importantly, we proved that healthcare AI does not need to sacrifice privacy in order to be useful.
What we learned
Through this project, we learned:
How Zero-Knowledge Proofs can solve real-world privacy problems Practical usage of Midnight Network and Compact contracts The importance of privacy-by-design architecture How to combine AI systems with cryptographic verification Secure backend design principles The growing need for trust in healthcare technology
We also learned that building impactful technology is not only about innovation, but about solving genuine human problems responsibly.
What's next for PrivaMed AI
We plan to transform PrivaMed AI from a hackathon prototype into a production-ready privacy healthcare ecosystem.
Future plans include:
Real deployment on Midnight mainnet Integration with wearable health devices Anonymous privacy wallet support Specialist referral systems Multi-language accessibility Mobile application development Telemedicine integrations Population health analytics using anonymized insights White-label SDK for healthcare companies
Our long-term vision is to make PrivaMed AI the foundational infrastructure for privacy-first healthcare intelligence worldwide.
Log in or sign up for Devpost to join the conversation.