Rust NYC: Zero Knowledge Proofs & GPUs in Rendering
New York City — United States

When
10/8/2026, 10:30:00 PM
Where
436 Lafayette, Floor 6, New York City, NY, New York City, NY, us
About
Join us on Thursday, October 8th at our sponsor Materialize. Doors open at 6:30 p.m. to give attendees plenty of time to grab pizza and socialize, and the talk begin at 7:15 p.m.We have two incredible speakers for y'all!
John Wu is a Research Engineer at Nethermind and will be speaking on Building Zero Knowledge Applications in Rust.
I work in cryptography and high-performance computing: my research is on making zero-knowledge proofs much faster, which turns out to be as much about data representation and memory behavior as it is about the mathematics. One thing that buys you: proving your identity directly from a government-issue id without having to share your personal information like name or address to other people. Before that I built security tooling for Apple's SRE team that deployed seamlessly onto Kubernetes across AWS and Apple's bare-metal machines, consensus and networking systems for Mina Protocol's zero-knowledge blockchain, and operator lowerings in LLVM's torch-mlir. I like ZK most when the math and the machine code have to agree.
This talk demystifies zero-knowledge proofs for Rust developers by breaking them down from first principles: arithmetic constraints, witnesses, finite-field polynomials, sum-check, commitments, and folding schemes. We'll explain how a computation becomes something a prover can convince a verifier about, without revealing the underlying data. Then we'll connect the math to real implementation work with a Rust case study: a small-value optimization I implemented in Microsoft's open-source prover, speeding up exactly the repetitive hashing and signature checks that a passport proof is built from, by letting small values stay small instead of paying full field width for every one. That's what makes the application real — proving your identity from a government passport in real time, rather than as an offline batch job. The goal is to leave attendees with both a conceptual map of ZK and a concrete feel for how high-performanc