The problem GitRaven solves

We are aiming to promote more opensource contributions by fresh opensource contributors through the power of blockchain and AI.
Gitraven is a platform where opensource project maintainers can list their projects and issues on chain to allow contributors to earn rewards like cryptocurrencies and NFTs and buidl up their contribution profile entirely on chain.

Features of our platform:

• Easy onboarding for non web3 users
• Tipping system for both contributors and maintainers
• Crypto and NFT collectibles for contributors
• Leaderboard system to gamify the platform
• Promote more users getting into web3 and opensource through an incentive reward system
• All data is on chain
• Insanely low gas fees (zkEVM supremacy 😎)
• AI enabled

We are utilising AI models to:

• Categorize issues into 3 different categories: 'bug', 'enhancement' and 'question' to match users with the type of issues they like the most.
• Use the stable diffusion model to generate images using keywords related to issues, this image will be used in minting NFTs as a reward.

Challenges we ran into

Initially we were using use the new blockchain tech called TableLand which can be used to create SQL tables on chain for easy data storage and retrival. However we later found out that Tableland doesnot support polygons zkEVM chain, which we wanted to use in order to minimize transaction fees. So we have to rewrite the entire contract to switch tableland with the contract state and had to use a number of mappings to achieve what we were doing with tableland.

Stable diffusion and NLP models are very resource intensive, so it was a challenge for us to optimise those models as much as we can to make up for the low processing power of repls. The NLP model had to be quantized as well as pruned just to be able to load it into repl.it