Vumah Grant Proposal

[Wilfychep]

Project Abstract

Please replace these instructions with a brief description of your project summarising key points (1-2 paragraphs).

If your application is a follow-up to a previous grant, please mention which one in the first line of the abstract and include a link to previous pull requests if applicable.

Grant level

• Level 1: Up to $10,000, 2 approvals
• Level 2: Up to $30,000, 3 approvals
• Level 3: Unlimited, 5 approvals (for >$100k: Web3 Foundation Council approval)

Application Checklist

• The application template has been copied and aptly renamed (project_name.md).
• I have read the application guidelines.
• Payment details have been provided (Polkadot AssetHub (USDC & DOT) address in the application and bank details via email, if applicable).
• I understand that an agreed upon percentage of each milestone will be paid in vested DOT, to the Polkadot address listed in the application.
• I am aware that, in order to receive a grant, I (and the entity I represent) have to successfully complete a KYC/KYB check.
• The software delivered for this grant will be released under an open-source license specified in the application.
• The initial PR contains only one commit (squash and force-push if needed).
• The grant will only be announced once the first milestone has been accepted (see the announcement guidelines).
• I prefer the discussion of this application to take place in a private Element/Matrix channel. My username is: @_______:matrix.org (change the homeserver if you use a different one)

[semuelle]

Thank you for the application, @Wilfychep, but please fix the Markdown syntax in your document. It is currently very hard to read and our automated checks rely on it.

[Wilfychep]

Hi @semuelle I have made the changes. Thanks.

[semuelle]

Thank you for updating the proposal, @Wilfychep. A few follow-up questions:

• Based on the team section, you have a lot of experience with AI and blockchain research and projects. Do you have any material about these (and any zk work you have done in the past) you could share, such as published articles, GitHub repos/profiles or LinkedIn profiles?
• You seem to have worked on a StarkNet prototype. Why are you shifting to Polkadot?
• Could you provide an architecture diagram or similar, so I can understand the parts of the system better? Where does the AI model live/execute?

[Wilfychep]

Thanks @semuelle for your questions.

1. Previous AI, Blockchain & ZK Work

The Vumah Labs team has a solid track record in AI, blockchain and zero-knowledge research, combining applied machine learning with decentralized trust architectures. Below are selected works and profiles relevant to our experience:

• Kweli – AI + ZK Deepfake Verification Prototype (Starknet, Cairo v2)

• Aleo Developer Advocacy (2023)

  • Conducted hands-on zkSNARK tutorials and developer workshops under Aleo Africa’s initiative.
  • Authored learning resources and circuits for privacy-preserving application design.

• Research Publications

  • Omar SM, Kimwele M, Olowolayemo A, Kaburu DM. Enhancing EEG Signals Classification using LSTM-CNN Architecture.
    Engineering Reports (2023) — DOI:10.1002/eng2.12827
  • Chepkwony W. (2024). AI, Zero Knowledge and the Future of Digital Trust in Africa.
    Published via Vumah Labs Research Series — https://medium.com
  • Media, A. & Maulid, S. ICT-Based Communication Channels for Knowledge Sharing among Multicultural Students.
    IGI Global (2012)

---

2. Why We’re Shifting from Starknet to Polkadot

Our Starknet prototype successfully validated AI + ZK proof verification. However, Polkadot offers stronger modularity, interoperability, and developer tooling to scale this vision into a verifiable media trust network.

Key motivations for the shift:

• Substrate Modularity: Enables development of a custom pallet (kweli_verification_pallet) for direct on-chain proof verification without relying on external L2 logic.
• Cross-Chain Proof Exchange: Using XCM and HRMP, the RailKit layer will allow proof attestations to flow between parachains (e.g. with KILT for identity and Frequency for content provenance).
• Developer Ecosystem: Polkadot’s mature SDKs (Substrate, Polkadot.js) allow more rapid developer adoption and verifiable interoperability.
• Scalability & Governance: Polkadot’s on-chain governance model supports transparent, upgradable verification logic aligned with ethical AI governance.

Our migration from Starknet is therefore not a pivot but an expansion building a broader, interoperable framework for ZK-based authenticity verification that integrates with the Polkadot ecosystem.

---

3. System Architecture Overview

The Vumah Labs system comprises three primary layers, plus an interoperability bridge (RailKit) designed for decentralized, cross-chain verification.

1. AI Inference & Proof Generation Layer (Off-chain)

   • Executes pre-trained EfficientNet-lite models for deepfake/media authenticity detection.
   • Runs in a containerized environment (Flask microservice with GPU support or cloud inference node).
   • Outputs an authenticity confidence score and generates a zero-knowledge proof of model output integrity.
   • The model executes off-chain to ensure flexibility and privacy-preserving computation, with results transmitted securely to the on-chain layer via API.

2. On-Chain Verification Layer (Substrate Runtime Pallet)

   • Implements kweli_verification_pallet that stores:
     • Media content hash
     • ZK proof hash
     • Verification result (true/false)
   • Validates the proof structure and provides immutable audit trails for media authenticity.
   • Provides extrinsics for proof submission, validation, and query.

3. Developer Integration Layer (Vumah SDK)

   • TypeScript SDK enabling developers and external dApps to interact with the pallet.
   • Exposes APIs for querying verification data, generating new proofs, and managing wallet-based verification submissions via Polkadot.js.

4. Interoperability Layer (RailKit)

   • Enables cross-parachain proof transfer and attestation synchronization using XCM.
   • Integrates with:
     • KILT: For decentralized identity linkage to proof ownership.
     • Frequency: For content broadcasting and provenance tracking.
   • Facilitates a multi-chain reputation system where verified media can be recognized across the Polkadot ecosystem.

---

4. Execution Context

The AI model executes off-chain within a secured containerized environment.

Proof generation and hash commitments are transmitted on-chain to the Substrate pallet for verification.

Cross-chain interoperability is handled through RailKit using XCM messaging

---

5.