Data science · Quantitative modelling · Applied machine learning
I build and validate mathematical models for decisions under uncertainty, and turn them into Python software. My background combines applied mathematics and data science with financial model development and validation in Chile and Canada.
My work includes credit-limit optimisation, real-estate haircut estimation, Bayesian expected credit loss modelling, inflation forecasting and NLP. I am particularly interested in the connection between mathematical formulation, uncertainty, numerical accuracy and useful implementations.
Python software for option-pricing PDEs using finite elements, with convergence checks, analytical price/Greek comparisons and documented capability limits. Recent work explores reduced-order models and experimental Bayesian and JAX integrations.
The repository's reduced-order Black–Scholes benchmark records a 28.6× median online speedup on synthetic holdout cases. This is a bounded experimental result: offline costs, accuracy gates and the tested parameter range are documented in the benchmark report.
Numerical option pricing and sensitivities, including Black–Scholes analytical checks, Rannacher smoothing and one-dimensional early-exercise reference cases.
A small educational application connecting a continuous linear-programming model to a Django interface with PuLP and CSV input/output.
- Bayesian inference, inverse problems and uncertainty quantification.
- Numerical PDEs, finite elements, finite differences and reduced-order modelling.
- Applied ML, time series, optimisation and stochastic simulation.
- Reproducible scientific software: explicit assumptions, tests and understandable limitations.
I work mainly in Python with NumPy, SciPy, pandas, Polars, scikit-learn and PyTorch. Recent project work also uses scikit-fem, pyMOR, PyMC/ArviZ and JAX/NumPyro.
Based in Santiago, Chile. Spanish, English and French.