STIR+
Notes on STIR-style proximity testing, where each round cuts the degree by k but only halves the domain: the construction, how to handle repeated queries, and round-by-round soundness.
Research Engineer · ZisK
I build systems that prove computation without revealing secrets.
I work on zero-knowledge proof systems, from the math to the prover code. Right now I'm building ZisK, a high-performance zkVM. Before that I was on the Polygon zkEVM proving stack, where I worked on the arithmetization and co-designed eSTARK, the proof system at its core. My origins are in mathematics and applied cryptography. Always looking for elegant solutions to hard problems.

Research engineer on ZisK, a high-performance zero-knowledge virtual machine for proving arbitrary programs.
A Lean 4 roadmap for formally verifying ZisK's recursive aggregation: specifying the STARK verifier in Lean, proving the verifier circuits sound, and showing that an accepted final proof yields valid, bus-balanced witnesses.
Generalized STARKs with the eAIR intermediate representation and three built-in arguments. The core proving system of the Polygon zkEVM prover.
PlonKup integrates plookup into PlonK. RapidUp extends lookups across multiple domains with a selector-based permutation protocol.
Notes on STIR-style proximity testing, where each round cuts the degree by k but only halves the domain: the construction, how to handle repeated queries, and round-by-round soundness.
Keep STARK DEEP quotients over the base field by dividing by the minimal polynomial of the out-of-domain point instead of X − z, so base-field columns never touch extension-field arithmetic.
Computing a STARK's out-of-domain evaluations quickly with barycentric interpolation over the trace subgroup.