Performances
RD Popularity Index
RD Growth Index
Comparison
Follow Updates
Follow Lists
People
About Fair Math
Fair Math is a cryptography infrastructure company whose core product, FHE Computer, is a decentralised execution environment where heterogeneous compute nodes run computation directly on ciphertext. It addresses the inability of sensitive data to enter AI training, inference and cloud analytics: data is encrypted locally before entering the system, stays encrypted throughout computation, and only authorised parties can decrypt results. The architecture spans application, orchestration, execution and network layers with an IPFS-based data layer, and the operating system borrows UNIX concepts of processes, threads and virtual memory for resource allocation and task scheduling.
Supporting components include the FHERMA challenges platform, the POLYCIRCUIT component repository and fhelanfg, an MLIR-based FHE compiler, with use cases covering encrypted vector database retrieval, cross-organisation joint computation and private inference. On the payments side it offers a modular privacy framework that selectively encrypts transaction amounts and balances without changing the underlying chain. The computational cost of ciphertext operations remains far higher than plaintext execution, and throughput depends on the size of the heterogeneous node network.
In 2026, Fair Math worked with NVIDIA to open two GPU-accelerated challenge tracks on its FHERMA platform, high-degree polynomial arithmetic and CKKS key switching, built on the cuPQC-NTT and cuPQC-BigInt libraries in the NVIDIA cuPQC SDK. The goal is to crowdsource reusable, high-performance FHE primitives and lower the performance barrier for homomorphic encryption in healthcare, finance and machine learning.
The product line includes the Fair Math Computer, a decentralized FHE execution environment using UNIX-like abstractions for applications, processes, threads and virtual memory with an IPFS-based data layer, plus the FHERMA challenge platform, the POLYCIRCUIT component repository and the MLIR-based fhelanfg compiler, with a command line interface. The team says its new-generation FHE evaluates amortized about 1,000 times faster than third- and fourth-generation schemes, supporting arbitrary functions and SIMD vectorized processing.
The Fair Math team will attend ACM CCS 2026 and WAHC 2026 in The Hague in November 2026 and NeurIPS 2026 in Sydney in December, followed by FHE.org and Real World Crypto in Seattle in April 2027 and IEEE S&P in Montreal in May 2027, pushing standardization and industry adoption of encrypted computing.
Gurgen Arakelov — Founder & CEO; PhD in mathematics and computer science; founded Fair Math in 2023 and leads FHERMA.
Elvira Kharisova — Co-founder & VP of Operations and Partnerships.
Andrey Nekrasov — CMO.
Valentina Kononova — Head of Artificial Intelligence.