Shelby Hernandez

About Me

I'm an ever-curious and creative thinker who enjoys finding innovative ways to solve complex problems. Since completing my M.S. in Applied Statistics, I've worked in Property & Casualty reserving, developing models that forecast loss emergence, quantify uncertainty, and guide financial strategy. Over time, my work in actuarial analytics has deepened my appreciation for how data, modeling, and scientific reasoning can intersect to better understand the world around us.
That curiosity has led me toward Catastrophe Risk Modeling, where statistical rigor meets the natural sciences. I';m drawn to how catastrophe models use data from seismology, meteorology, and hydrology to simulate real-world events and quantify their financial impacts. It's a space that allows me to apply my technical and actuarial background in loss reserving frameworks, stochastic simulation, scenario analysis, and model validation, while expanding into geospatial modeling and hazard analysis.

I'm currently building a portfolio of catastrophe models across earthquake, flood, and hurricane perils, integrating open-source tools like OpenQuake, ArcGIS, and Python. Through this work, I'm combining my analytical training with a lifelong fascination for natural phenomena, aiming to build models that not only predict losses but also help communities, insurers, and policymakers prepare for the risks ahead.

Cat Model Portfolio

Earthquake (CA)

Flood (Gulf of Mexico)

Hurricane (East Coast)

Probability & Stochastic Modeling

Poisson processes, survival models, expected value & variance derivations, limited expected value, joint distributions

Modeling event frequency and severity, e.g. earthquake occurrence rates, flood peaks, hurricane landfall frequency

Statistical Inference & Data Analysis

Hypothesis testing, parameter estimation (MLE, unbiasedness, efficiency), handling truncation & censoring, model validation

Estimating loss distributions, fitting vulnerability or fragility curves, validating hazard vs. observed loss data

Extended Linear Models (GLMs & Beyond)

Link functions, deviance analysis, variable selection, model diagnostics, AIC/BIC, control & offset variables

Building intensity-damage or loss models, rate-relativities, and exposure adjustments across geographic zones

Credibility & Bayesian Methods

Bühlmann, Bühlmann-Straub, Bayesian credibility frameworks

Weighting simulated vs. observed losses, blending regional hazard data with local claims experience

Linear & Mixed-Effects Models

Hierarchical modeling, random effects, correlated error structures

Capturing correlation in multi-region loss data or repeated-event simulations (e.g., annual aggregate losses)

Statistical Learning & Machine Learning

K-nearest neighbors, decision trees, ensembles, PCA, clustering, neural networks, predictive accuracy metrics (Lift, Gini, AUROC)

Classifying exposure types, clustering geographic risk zones, or predicting damage classes from hazard intensity

Time Series & Forecasting

ARIMA models, trend & seasonality decomposition

Modeling temporal patterns of insured losses or climate-driven hazard frequencies

Actuarial & Financial Applications

Loss reserving frameworks, stochastic simulation, scenario analysis, exposure modeling, reinsurance structuring

Integrating financial terms (deductibles, limits, reinsurance) with catastrophe loss outputs

Programming & Tools

Python (pandas, NumPy, SciPy, Matplotlib), R, SQL, Excel/VBA, ArcGIS Pro, OpenQuake, QGIS

End-to-end model implementation, spatial data processing, loss simulation, and visualization

Communication & Collaboration

Translating quantitative models into actionable insights, visualization, report writing, stakeholder communication

Explaining uncertainty and risk implications to underwriters, actuaries, and decision-makers