Nesa provides access to a catalog of pre-trained machine learning models for various AI tasks. These models cover capabilities such as text classification, content summarization, image generation, and language translation. The platform enables users to deploy these specialized models for immediate application in their workflows.
Funding
Funding not disclosed
Founders
Product
Problem
Current AI inference solutions often lack privacy, making them unsuitable for applications involving sensitive data in sectors like healthcare and finance. Existing centralized AI platforms also present risks related to data ownership and control.
Solution
Nesa is a Layer-1 blockchain network designed to bring AI on-chain with a focus on privacy and trust. The Nesa platform allows applications and protocols to seamlessly integrate with AI models while ensuring data confidentiality through technologies like Zero-Knowledge Proofs (ZKPs) and secure multi-party computation (MPC). Nesa utilizes the AI Terminal (AIT), an end-to-end execution interface and machine network for running inference queries on-chain. The platform containerizes AI models and query templates, hosts an ecosystem of off-chain services, and provides a network of miners for AI inference execution.
Target Audience
Nesa targets AI developers, Web3 developers, and enterprises seeking a secure and decentralized platform for AI inference, particularly those working with sensitive data or requiring verifiable AI computations.
Features
- Fully decentralized AI inference network leveraging a Layer-1 blockchain.
- AI Terminal (AIT) for end-to-end execution of inference queries on-chain.
- Support for over 100,000 AI models across 29 modalities.
- Privacy-preserving computation using SMPC and ZKP schemes.
- Trusted execution environments (TEEs) for secure secret share distribution.
- Standardized and secure execution environment for containerized AI models.
- Inference Request Queueing for high throughput and low latency execution.
- Fair and secure inference committee selection using VRF.
- Hybrid enhanced ZK privacy system with a two-phase transaction structure.
- Neural Arbiter Network (NAN) for kernel validation testing.
- Native NES token for gas settlement, staking, and miner/model owner rewards.