ezkl

Zero-Knowledge Proof for Nuclear Disarmament Verification

A deliberate affront to egalitarian tokenomics—cryptographic proof that not all wallets are created equal. What can be claimed without reputation can be excluded without apology.

This project implements the proof-of-concept of blockchain-based federated learning system with zero-knowledge proofs for model aggregation verification. MSc thesis in CS University of Tartu.

ZKML (EZKL) proofs over fNIRS BCI models — mental arithmetic vs baseline (ZuThailand EZKL Prize). Colab in ~5 min.

Documentation Repository for EZKL pipeline

Benchmark-driven zkML DeFi oracle comparing compiler-generated EZKL and hand-written Circom circuits for LoRA-adapted risk inference, with EVM verification and reproducible benchmarks.

A Proof of Concept for a Zero-Knowledge Latent Bridge. Demonstrates trustless, on-chain EVM verification of off-chain PyTorch/ONNX machine learning transformations using EZKL and Foundry

Code for collecting ezkl performance metrics on Zero-Knowledge Proof generation over inference tasks for multiple deep neural network models.

Supply-chain security for AI fine-tuning. This repo generates zk-SNARK proofs per gradient step, binds proving keys to TEEs, prevents rollback, and publishes audit artifacts for public replay

zkML-powered liquidation protection for Aave V3 using Halo2, EZKL, and ERC-7579 smart accounts.

High-performance physics-based verification engine for neural identity and trustless ZK-observability.