AI for Math & Pure Mathematics Research
As an active researcher in the emerging field of AI for Math, my primary ambition is to engineer fully automated pipelines for mathematical discovery. This involves generating novel conjectures, unifying formulas for fundamental mathematical constants, and developing automated reasoning and theorem-proving systems.
My theoretical work bridges pure mathematics and algorithmic exploration. I study the geometry of pFq parameter spaces through p-adic arithmetic and solve number-theoretical problems concerning Conservative Matrix Fields (CMFs) utilizing algebraic geometry.
The Euler 2 AI Project Operating as an independent researcher collaborating with the Ramanujan Machine team in Israel, I am heavily involved in the ”Euler 2 AI” initiative. This work focuses on neuro-symbolic theory exploration and the unification of zeta functions, primarily targeting ζ(3) and ζ(2n+1).
2026 EuroHPC LUMI Project: Ramanujan Dreams To push the boundaries of automated conjecture generation, I am driving a massive computational initiative: ”Ramanujan Dreams on EuroHPC LUMI: Large-Scale Hypergeometric-to-CMF Exploration for ζ(3), ζ(5) & ζ(2n+1).”
Supported by an allocation of 90,000 GPU hours on the LUMI-G supercomputer in Finland, this project aims to address a profound open question in number theory: How many distinct Conservative Matrix Fields and continued fraction families correspond to constants like ζ(3), ζ(5), e, and ln2?
Our methodology relies on:
- Anchor Seeds: We utilize classical Apéry-type hypergeometric identities (such as the 4F3 family at z=1) and convert them into CMFs whose induced numeric sequences converge rapidly to ζ(3).
- Large-Scale Sweeps: Using these anchor seeds, we generate thousands to millions of contiguous and shifted CMF variations, escalating the most mathematically promising cases for deeper verification.
- HPC Acceleration: By combining this seed-and-variation model with GPU-accelerated linear algebra and MPI ”rank-per-GPU” parallelism, we systematically sweep the CMF landscape at an unprecedented scale.
The ultimate objective is to map the underlying ”structure of constants,” identifying how these families relate through transformations and parameter shifts to build a comprehensive atlas of constant-CMF relationships.