Solutions provided by Deuro
Here we propose our innovative AI solutions based on the blockchain to solve the problems mentioned above.
gStorage: For the data security and privacy problem, we propose the gStorage solution. With AI special use case and blockchain smart contract support in mind, we introduced the distributed end-to-end encrypted solution to store/retrieve the training data, the production prediction data, and the model files.
gCrawl: For the lack of the labeled data problem, we propose the innovative gCrawl architecture. The gCrawl system is composed of fault-tolerant distributed high-performance data crawling system, stateful NoSQL database(deduplication as well as metadata storage), decentralized storage, transfer learning for model parameters to finish the feedback loop.
gPredict: For the algorithm accuracy problem, we proposed an innovative way to train/evaluate/test the model performance on the blockchain. gPredict includes floating point mapping, hyper-parameters routing, and decoding results reducing.
gCompute: For the high hardware as well software costs when developing, testing and deploying AI systems, we introduced the gCompute module. The gCompute module is scalable, controllable, difficulty, maintainable, responsive and concurrent.
1. gStorage: Privacy-Preseving Decentralized Artificial Intelligence client-driven data storage system
In order to securely store the user recording in a trustless way, we have to ensure privacy and security for our customer. Here we propose an innovate way to store user data on the blockchain. Below is the workflow chart for how gStorage works.