On modeling compute markets
Why compute should be modeled as term structures rather than a single price, and what that buys for hedging, reservation pricing, and collateral.
I do research on mathematical finance, mainly on volatility models. Currently based in Hong Kong, otherwise in Barcelona. Tennis when I can.
Writing
Projects
Recursive Discovery
A compact system for recursive mathematical and empirical scientific discovery.
Compute Market Model
Stochastic delivery curves for compute pricing, infrastructure financing and portfolio control.
Research
Volterra-Perron Pricing of the Option Manifold
Formulates option surfaces as the public quotient of a latent Volterra-Perron state, deriving recovery, residual-pricing, native-Greek, synchronization, and finite-benchmark results.
Synchronized Stress Relative Value Public-Neutral Option Books under Latent Stress Compression
Builds a public-neutral listed-option rule that sells index synchronization convexity against carrier-basket convexity only after an adapted lower-bound sellability gate is positive.
Perron Stress Pricing under Public Compression Residual Price Intervals, Synchronized Stress, and Visible Hedging
Develops residual pricing, stress-charge selection, synchronized carrier absorption, and visible-hedge decomposition for Perron stress under public compression.
Risk-Neutral Compression and the Geometry of Latent Market Stress
Formulates risk-neutral compression: public option surfaces observe conditional projections of latent Perron loadings, with SPX/NDX evidence from matched options and realized-volatility data. Companion note: Gaussian Seams for Risk-Neutral Compression: A Volterra Source-Family Companion.
A Rough Theory of Markets Latent Contagion, Market Phases, and Market-Variance Risk
Derives market roughness, Perron synchronization, jump-like coarse observations, and incomplete visible hedging from directed Gaussian Volterra contagion.
Detecting Latent Volatility Contagion A Projected Score Estimator for Rough Volatility Models
Constructs a projected covariance-score estimator for source-screened latent contagion; Oxford-Man evidence yields dense directed maps read through intensity and stability.
Dynamic Observability of Latent Contagion Sequential Revelation and Observable Screening
Distinguishes among finite-resolution evidence, oracle prediction, and observable-history prediction, showing how dynamic contagion can survive path tests while being screened in feasible forecasts.
Latent Volatility Contagion in Rough Volatility Models Spectral Thresholds and Detectability
Separates option-skew visibility from path-law detectability for directed rough-volatility contagion, identifying the latent regime between the two Gaussian Volterra thresholds.
Last updated: