Research areas: scientific computings, financial systems and growth, deep uncertainty quantification, big (complex) data analysis

New Papers

Mortality Prediction as Boundary Value Problem, February 2026, DOI: 10.5281/zenodo.18841509

The HJB/viscosity framework that revolutionized macro (Achdou-Lasry-Lions, Kaplan-Moll-Violante, and the HANK literature that followed) applies to predicting mortality from medical event sequences.

Mortality prediction reframed as Hamilton-Jacobi PDE on patient trajectory space. Discrete medical events lift to rough path space via piecewise-constant embedding. Transformer attention computes signature projections. ReLU networks yield viscosity solutions with tropical hypersurface decision boundaries. Death as absorbing boundary condition.

The intellectual lineage traces to work with Ben Moll and Rob Townsend on heterogeneous agent dynamics in economic development (NBER WP 19618, PNAS 2017). The PDE machinery differs; the foundational commitment is the same. Representative agents destroy signal in macro. Feature vectors destroy signal in medical prediction. Heterogeneity carries information - in wealth distributions, in patient trajectories. Collapse it and you lose what matters.

One architecture. <1M parameters. Outperforms best academia/industry approaches at a fraction of the complexity, fully explainable, easily scalable on Intel compute, no GPUs required.

Why Financial Event Sequences Require Temporal Transformers: Necessity Theorems for Default Prediction, January 2026, Available at SSRN: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6154026 , DOI: 10.13140/RG.2.2.35041.10084

The theoretical framework identifies minimal requirements for financial sequence modeling and explains observed performance gaps between sequential and tabular approaches in credit scoring.

Causal Set Topology for Transformer Architectures: From Planck Scale to Human Life, December 2025, DOI: 10.5281/ZENODO.18449228

Mathematical unification connecting causal set theory in quantum gravity to temporal modeling of human trajectories. The same tensor operations describing discrete spacetime describe a patient evolving through clinical events.