R Metropolis sampler for 10-factor portfolio VaR/ES, benchmarked against the exact posterior to show how poor mixing understates parameter risk.
Survival analysis of CISA KEV listings: how likely and how fast CVEs are listed, and which patch SLA design leaves least exposure.
An absorbing Markov chain model of the cyber kill chain in R, quantifying dwell time and containment odds under baseline and hardened defenses.
A pure-R detection system pairs PageRank-based structural risk mapping with an unsupervised autoencoder and Extreme Value Theory thresholds to catch stealth cyber attacks.
Applies Google’s PageRank algorithm to a 25-node enterprise IT dependency graph, ranking structural single points of failure for board-level risk budgeting.
ATLAS-based FRR/FAR/EER threshold calibration for detection rule: framing SOC alert tuning as a documented risk decision for cybersecurity leaders.
Which node’s failure breaks the most? Graph centrality and node-removal simulation applied to a 90-node fintech infrastructure topology to rank structural risk and surface the failures that conventional inventories miss.
Finds a ~30% capital benefit from modeling operational risk dependence with copulas, and shows the result barely depends on which copula is chosen.
Quantifies how much a log-normal capital model understates operational loss tail risk versus a Generalized Pareto fit, using NexaCore’s loss register.
An identity governance case study demonstrating how graph theory surfaces hidden privilege escalation paths in enterprise Active Directory environments that static access reviews cannot detect.
A dependency graph model showing 8 ICT vendor entries in a DORA Register of Information collapse to 5 failure domains for a EU asset manager. Calibrated to the ESAs’ 19 designated CTPPs, a discrete-event simulation produces an annualized business interruption loss distribution.
Most fintech risk registers treat ransomware as a qualitative category. This model replaces the label with a loss curve — a full distribution of probable annual financial outcomes, expressed in dollars, with Value-at-Risk figures a CFO can act on.
This project introduces two R packages — hfhub and tok — and applies them to a structured metadata audit of FinBERT (ProsusAI/finbert), a BERT-based sentiment classifier widely deployed in fintech AI pipelines.
This project introduces the Policy as Code concept through the simplest possible worked example — a six-condition password complexity validator — to make the pattern legible to both compliance professionals and engineers.
Survival analysis exposes where RTO targets break down — and why your mean recovery time is lying.
Analyzing the digital fingerprints of URLs through machine learning and the lightGBM model to detect incoming phishing emails.
Bayesian Kelly position sizing: replaces the point estimate with a Beta posterior, derives the full Kelly distribution across four scenarios, and builds a staking policy from estimation uncertainty.
Bayesian lending risk analysis in R using rstanarm & bayestestR — posterior distributions, credible intervals, ROPE, and pd applied to real-world Lending Club interest rate data.
A comprehensive data science framework leveraging machine learning and survival analysis for identifying, analyzing, and predicting customer attrition within a consumer credit card division.
Prediction analysis of CVE exploitation using Machine Learning. Integrates NVD, EPSS, and CISA KEV data to forecast active threats. Features Random Forest modeling, SMOTE balancing, and executive risk visualization to enable proactive vulnerability prioritization.
A technical framework and narrative summary project for implementing Block Bootstrap and EVT in R. It provides a practical demonstration of how to quantify tail risk in high-volatility assets (specifically Bitcoin) using advanced statistical methods that outperform standard Gaussian models.
An interactive R Shiny app modeling the Defender’s Dilemma via Markov Chains & Monte Carlo simulations. Visualizes the stochastic nature of resource depletion (SOC fatigue, budget) vs. infinite attacks, calculating a specific “Probability of Ruin” to guide risk management.
An Interactive R Shiny app simulating Gambler’s Ruin via Monte Carlo methods. Visualizes Markov Chain dynamics to demonstrate how finite resources face inevitable ruin against infinite odds. Features dynamic inputs for capital & win rates to illustrate risk in trading & cybersecurity.