This repository is a curated collection of resources for Zero-Knowledge Machine Learning (ZKML), targeting researchers, developers, and enthusiasts interested in combining ZK proofs with AI/ML. It provides a comprehensive overview of the ZKML landscape, including papers, codebases, community calls, and use cases, aiming to foster innovation and adoption in this emerging field.

How It Works

ZKML leverages zero-knowledge proofs to enable verifiable and private execution of machine learning models. This allows computations, such as AI model inferences, to be proven correct and private without revealing the underlying data or model parameters. The repository highlights various approaches, including specialized ZK-SNARK circuits for ML tasks and general-purpose ZK virtual machines, facilitating trustless AI integration into blockchain and other privacy-sensitive applications.

Quick Start & Requirements

This is a curated list, not a runnable project. To explore specific codebases, refer to individual project repositories linked within the README. Requirements vary per project, often including Python, ZK-specific libraries (like Circom, Plonky2, RISC Zero), and potentially GPU acceleration for ML tasks.

Highlighted Details

• Comprehensive list of ZKML papers, articles, podcasts, and community calls.
• Curated codebases showcasing ZKML applications like on-chain trading bots, MNIST inference, and decision tree predictions.
• Details on key ZKML projects and companies (e.g., Modulus Labs, ezkl, Risc Zero, Worldcoin).
• Exploration of ZKML use cases in privacy-preserving inference, computational integrity, and decentralized AI.

Maintenance & Community

The repository is community-driven with an emphasis on contributions. It links to ZKML community calls held bi-weekly, with recordings available on YouTube and notes provided. Key community hubs include a Telegram channel.

Licensing & Compatibility

The repository itself is not licensed as a software project. Individual codebases and papers linked within have their own licenses, which must be reviewed for compatibility, especially for commercial use.

Limitations & Caveats

This is a resource aggregator, not a unified framework. Users must navigate and integrate individual ZKML tools and libraries, which may have varying levels of maturity, documentation, and compatibility. The field is rapidly evolving, with ongoing research and development.