Independent Research Initiative • Taiwan • 2026

Medical AI Agent ZK Trust Layer.

Verivis Lab is building a trust layer for medical AI agents: a research-stage architecture for inference attestation, onchain anchoring, and verifiable audit trails in high-stakes clinical workflows.

Our current implementation focus is now aligned with the Base ecosystem. The long-term goal is to make medical AI workflows more auditable, more accountable, and easier to evaluate across hospitals, regulators, and machine-native systems.

Research notice: Verivis Lab is an independent research initiative and is currently not a registered legal entity. The materials on this website describe research concepts, prototype interfaces, and ongoing scholarly work. Nothing on this website is intended as medical advice, clinical diagnosis, treatment guidance, or a regulated medical device claim.

Try Live Demo Read Latest Update Contact the Founder

Verivis Lab visual for medical AI trust infrastructure

News / Updates

Latest Build and Directional Milestones

Reusable update module

New Strategic Direction • Base Ecosystem

Verivis Lab is now building on Base as its primary chain focus

Verivis Lab has formally transitioned its active chain focus from Solana to Base (Coinbase L2). This shift reflects the lab’s updated product direction around a Medical AI Agent ZK Trust Layer and a more EVM-native trust and verification stack.

Current Priority

Base-aligned Medical AI Agent trust infrastructure

Current Demo Network

Base Sepolia

Core Product Position

Medical AI Agent ZK Trust Layer

Why It Matters

Current prototype work is now centered on Base-native contract calls, trust anchoring, and EVM-compatible verification flows.

Note: Solana work remains part of Verivis Lab’s historical prototype journey, but Base is now the active implementation priority.

Archive Historical Milestone • Solana Devnet

Historical record: Anchor contract deployed to Solana Devnet

Before Verivis Lab shifted its implementation focus to Base, the project successfully deployed an Anchor contract to Solana Devnet. This remains an important historical prototype milestone in the lab’s early architecture exploration.

Program ID

6hWiksWjhRyopvyszzv4tLLzShJadNTcJG2nZ4mHipc1

GitHub Repository

github.com/petershih5/verivis-poi-mvp

Deployment Transaction

View on Solscan (Devnet)

Contract Address

View on Solscan

Core Positioning

Medical AI Agent ZK Trust Layer

Verivis Lab is now positioning its product architecture around a trust layer for medical AI agents: verifiable inference records, agent-level accountability, and cryptographically anchored auditability.

Current Chain Focus

Base / Coinbase L2

Current prototype deployment and contract interaction flows are now focused on Base, with Base Sepolia used for the live demonstration environment.

Clinical Domain

Medical AI, Nuclear Medicine, Trust

Verivis Lab explores how trust infrastructure can support high-stakes AI workflows in medical settings, especially where accountability and system boundaries matter.

Whitepaper

Strategic Whitepaper Overview

Version 2026.03 • Updated framing

A research framework for trust, verification, and climate-aware systems design in medical AI.

Verivis Lab’s architecture now centers more explicitly on a Medical AI Agent ZK Trust Layer: a system-level approach to inference anchoring, verification signaling, and accountable AI agent workflows.

Current implementation focus is Base-oriented. Earlier Solana work remains part of the project’s historical prototype record, not the active chain priority.

Open whitepaper summary

Executive Summary. Verivis Lab investigates how trust, sustainability, and cryptographic verification may be combined into a coherent infrastructure layer for medical AI research.

Medical AI Agent ZK Trust Layer. The current product framing emphasizes agent-level trust, onchain attestations, and verifiable execution records.

Invisible Risk Trust Framework. A research framework for privacy-preserving evaluation, expert accountability concepts, and measurable sustainability considerations.

Proof of Inference. A protocol concept for binding model execution claims to verifiable evidence without exposing sensitive inputs or proprietary internals.

Implementation Direction. Current prototype alignment is Base-focused, while prior Solana milestones are preserved as part of the project’s research history.

Research

Current Research Portfolio

Under Review

The Invisible Risk Trust Framework

Target venue: Frontiers in Blockchain
Author: Pei Fan Shih, MD

Submitted

Proof of Inference

Target venue: npj Digital Medicine
Author: Pei Fan Shih, MD

Submitted

From ALARA to ALACA

Target venue: Insights into Imaging
Author: Pei Fan Shih, MD

Submitted

Clinical Model Autophagy

Target venue: JMIR Medical Informatics
Author: Pei Fan Shih, MD

Publication note: No manuscripts, abstracts, or supplementary materials are publicly released on this website at this stage. Public links will be added only when release is appropriate.

Live Demo

Medical AI Agent ZK Trust Layer — Base Demo

Base Sepolia prototype

Prototype boundary: This demo is a research-stage interface. It hashes the uploaded example image client-side, anchors that hash through the Base Sepolia contract flow, and checks verification status onchain. It does not generate a full medical zk-SNARK proof in-browser, and it is not a clinical product or medical device.

Wallet and gas notice: This live demo requires the user to connect an EVM wallet, switch to Base Sepolia, and approve a testnet transaction. The connected wallet pays the Base Sepolia testnet gas fee for that transaction. This does not use Base mainnet funds, but it does require the connected wallet to hold enough test ETH on Base Sepolia.

Live Prototype Base Sepolia

Medical AI Agent ZK Trust Layer

Upload a non-identifiable example image to simulate a medical AI trust-layer workflow. The current prototype demonstrates wallet connection, Base Sepolia contract interaction, file-hash anchoring, and verification status feedback.

To complete the onchain step, the user must connect a wallet and confirm a Base Sepolia testnet transaction. Gas is paid by the connected wallet on testnet.

Current Contract

0x9DbB950EBc3fcbEfdfE320df07A8f995d0C0E976

Network: Base Sepolia

Upload Example Image

Select a non-identifiable PET/CT-style image for the prototype flow.

Approve Testnet Transaction

The connected wallet switches to Base Sepolia and pays testnet gas to submit the verification anchor transaction.

Read Verification State

The demo checks contract state and returns a trust-layer status onchain.

Input

Example image upload

⬆ Click or drag an example PET/CT-style image JPG / PNG • non-identifiable example images only • max 10 MB

Preview Local file only

Calling Base contract and awaiting confirmation...

Please confirm the Base Sepolia testnet transaction in your wallet. Testnet gas is paid by the connected wallet.

Output

Verification console

◇

Awaiting verification request

Upload an example image, connect a wallet, and approve a Base Sepolia testnet transaction to populate this verification console.

Verification complete

Medical AI Agent trust record has been anchored successfully.

Execution status

Confirmed

Environment

Base Sepolia

Contract address

0x9DbB950EBc3fcbEfdfE320df07A8f995d0C0E976

Derived file hash

—

Prototype trust-layer note

This demo anchors a deterministic hash of the uploaded file to the contract and then checks onchain verification state. The user must approve the Base Sepolia testnet transaction in their own wallet, and the connected wallet pays the required testnet gas.

Demo FAQ

Before you try the live demo

Quick answers for reviewers and first-time users

Do I need a wallet?

Yes. The current live demo requires an EVM wallet such as MetaMask because the verification request is submitted as a real Base Sepolia testnet transaction.

Do I need to pay gas?

Yes, but only on Base Sepolia testnet. The connected wallet pays the required testnet gas fee. This is not Base mainnet and does not use mainnet funds.

Will it use my own wallet balance?

It will only use the wallet that is currently connected and only after that wallet owner explicitly approves the transaction. Nothing is charged automatically without wallet approval.

What kind of image should I upload?

Only upload a non-identifiable example image. Do not upload patient-identifiable information, protected health information, or any confidential medical data.

Is this a medical device or diagnosis tool?

No. This is a research-stage prototype for demonstrating a medical AI trust-layer workflow. It is not a clinical product, not a medical device, and not medical advice.

What is actually being verified here?

The current prototype demonstrates a trust-layer flow: file hashing, wallet-based submission, onchain anchoring, and verification-state lookup on Base Sepolia. It should be interpreted as an early infrastructure demo, not a finished clinical verification engine.

Roadmap

2026–2027 Academic Roadmap

Q1 2026

Manuscript Publications

Targeted scholarly submissions across medical imaging, digital medicine, and trust infrastructure-related venues.

Q2 2026

ZK Trust Layer Demo Refinement

Refinement of public-facing Base-aligned demo flows and verification UX.

Q3 2026

Hospital Pilot Exploration

Exploratory conversations around retrospective, de-identified evaluation pathways where feasible.

Q4 2026 – 2027

Collaborative Clinical Studies

Longer-term ambition for academically grounded validation and partnership development.

Contact

Building Verivis Lab in public.

Verivis Lab welcomes research dialogue, technically serious collaborators, and institutions interested in the future of trustworthy medical AI.

We are actively looking to bring in more exceptional people across technical, research, and interdisciplinary roles. We especially welcome builders with strengths in software engineering, smart contracts, ZK systems, product design, medical AI, clinical informatics, nuclear medicine, and data infrastructure.

If you believe in the long-term need for a Medical AI Agent ZK Trust Layer and want to help shape Verivis Lab at an early stage, we would be glad to hear from you.

𝕏 GitHub

Founder: Pei Fan Shih, MD

Role: Independent Researcher

Focus: Medical AI, nuclear medicine, trust infrastructure, web3 systems

Location: Taiwan

Email: petershih5@gmail.com

Open to hearing from:

Technical builders and engineers
Medical AI researchers and practitioners
Smart contract / ZK / onchain infrastructure talent
Clinical and nuclear medicine collaborators
Mission-aligned interdisciplinary talent

Contact the Founder View GitHub