Compute.fi

• 

  DAO, Defi yield, dNFT and Compute(AI inference, training) architecture

• 

Our inspiration

The question was: "How can i pay for the use of this ai for a group of users without selling my crypto?" We like opensource ai development and we like crypto. If we create an ai workforce for our businesses or need ai inference for our applications we don't like to sell our crypto when its not an ideal time. When your crypto assets are waiting to be used they can generate yield. What is more valuable to us? yield or compute? If the cost of compute is going down and the value of the ai models increases we might generate more of a return with a working force of ai than simply collect a 5% yield on our eth treasury. If we automate the process of transmuting monetary energy in to computational energy we can use this as a reliable, verifiable building block for more complex systems.

Its nature, what it does.

The Project aims to create a decentralized funding mechanism for AI inference,training and compute where users or entire communities can operate open source AI models by using yield generated from their staked capital.

Construction

We divided the app in four parts. Defi, compute, dao and nft. One for each developer in our team. We started by laying out the architecture for the system and we made lots of diagrams. Each of us started building our respected parts. We coordinated closely between each other to synchronize our work.

The learning process

It has been a big learning experience to start the process of creating a application from an idea. We all got to push the boundaries of our abilities. We chose to build the most interesting concept we had even if it seemed to much to finish in this timespan. I don't think anybody in the team regrets that decision. We aimed to create a proof for ourselves that the idea is a feasible one where we could move on to create something that has not been done before. By building this project we learned how we can build it again, better.

Specific challenges

• working async in different time zones
• the busy live of teammates
• going from an idea to a design for a working system
• finding out the difference between: what we want to build and what we can build right now
• accepting to cut features because of time constraints
• looking for the perfect tech solution and finding the limitations of the current state of it

What we have accomplished

• we developed a idea we have not heard from before
• we designed a system that uses 3 main smart contracts to combine defi yield, NFT identification and compute calls
• we laid the foundation for a social aspect to the funding and sharing of ai inference trough a decentralized system
• our team created a blueprint for the yield to compute concept
• we researched the technology within blockchain that involves decentralized compute for ai applications and we found our candidate solutions to further develop our app in the future

Whats next for us?

We are going to develop more until we have a mvp. Then we will try to find validation within the community. Then we will use yield to compute to build fun products.

Built With

• chainlink
• docker
• dynamicnft
• external-adapters
• foundry
• hosting
• kaggle
• postgresql
• python
• shell
• solidity
• typescript