Hexs

Distributed ownership of AI models to offer fair incentives to open-source developers through chain-driven training.

Comment

Inspiration

An open-source developer usually doesn't have streamlined options to earn incentives while they contribute and with emerging AI, they are needed with deployment infra as well.

What it does

  • When the user tries to deploy an AI agent which he created on Akash, a smart contract instance for this activity is created.

  • This instance sets up an NFT minting smart contract which is used to encrypt the SDL file containing private docker image and CPU requirements.

  • Only those users with access to these can decrypt and re-deploy or download to modify the AI model.

  • By tokenizing these files, developers will be able to offer several more infra conditions by adding smart contract plugins like that of Soul-Bound NFTs.

  • Anonymous developers can fork other public models which they can fine-tune to provide better results.

  • With this mechanism, only those models with better generation will attract the most users.

  • The initial creators of the models whose forks are trending out will be generating royalties.

  • [Future] Each Model trainee will be able to generate API endpoints with which they can set API pricing and let other developers use their model.

  • Developers building out applications around these models will be generating revenues.

How we built it

  • Polygon Amoy for deploying smart contracts which tokenized infra. Contract Address

  • Avalanche shown as a CCIP receiver, supports all features for tokenizing infra. Contract Address

  • Scroll for deploying smart contracts which tokenized infra. Contract Address

  • Chainlink is being used in 4 ways:

  • Chainlink for bridging tokenized infra across multiple chains using CCIP. tx hash

  • Chainlink functions for validating user engagement on each of the Deployed Infra Models. tx hash (updates engagement score on-chain from Python Server which returns string value)

  • Chainlink Automation for automatically calling API functions to update all models' engagement scores on-chain. upkeep

  • Chainlink Data Feeds to translate prices between bridges on the marketplace. contract_address

  • Lighthouse to provide the feature for encrypting the SDL files before uploading to on-chain. github_loc

  • We are using the fork of Cloudmos to initialize our idea

  • Smart contracts were introduced to provide a set of decentralized features like Marketplace and DAO.

  • Python server for APIs. github_loc

Challenges we ran into

Running Akash blockchain's sandbox environment on the local machine was the most difficult part of building the application.

Accomplishments that we're proud of

We were able to construct a demonstration of how we can encourage developers to contribute by tokenizing infra and hosting each project's incentivization mechanisms.

What we learned

Learned in depth about Chainlink features and Akash blockchain.

What's next for Hexs

We've thought of our few future milestones:

  • Introduce burner wallets and gas sponsoring for a user-friendly interface

  • Transfer provider lease by ending with one user and starting with another as soon as the marketplace order is placed.

  • To implement a backend for developers to directly form APIs with their deployed models which they can offer to other devs for usage.

  • To implement SDKs for having extended conditions over models

Built With

Share this project:

Updates