Zero Knowledge Proofs(ZKPs) are increasingly used in the domain of privacy-preserving machine learning. However, recent works have not been widely incorporated due to the complexities of ZKPs, and they lack generalization. To address these issues, we present develop zkVML: Zero Knowledge Verifiable Machine Learning, a novel Python library that seamlessly integrates zero-knowledge proofs (ZKPs), allowing the verification of public inference over a private model. zkVML ensures privacy in both on-chain verification and off-chain computations, particularly within decentralized environments like blockchains. We design and develop several APIs, for handling complex computations present in Machine Learning models. We successfully validated our library by developing K-Nearest Neighbors(KNN), Decision Trees(DT) and Convolutional Neural Networks(CNN) models. While zkVML may incur increased execution times for complex models and large datasets, it offers significant privacy improvements alongside enhanced scalability, generalization, and control over model creation. Through high-level Python APIs abstracting complex ZKP functionalities, zkVML empowers developers to maintain flexibility and privacy control while guaranteeing data integrity. Our key contributions include generalized proof generation, user-friendly Python abstractions for easy integration, and on-chain verification via smart contracts. Future work may focus on optimizing proof generation times and extending zkVML’s functionality to a wider range of algorithms.