Nesa is a decentralized AI infrastructure platform designed for secure and private on-chain AI inference. It leverages advanced methods like zero-knowledge machine learning (ZKML) and split learning to ensure data privacy and trust. Nesa operates a decentralized query marketplace, where AI models are stored and accessed. Its hybrid sharding approach distributes computational loads across multiple nodes, making it accessible for nodes with varying resources. Nesa aims to overcome the limitations of centralized AI platforms by providing an open, secure, and efficient AI ecosystem.