risk-modeling

Experimental scenario analysis for real-life events forecasting with Codex or Claude

A research-grade lab for stress-testing DeFi protocols using Solidity mini-systems, a Python simulation engine, and a Streamlit dashboard. Simulates price crashes, liquidity shifts, AMM behavior, lending liquidations, and systemic risk dynamics. Designed for DeFi engineers, auditors, and researchers.

A research-grade tool that analyzes Solidity smart contracts for economic vulnerabilities such as unbounded minting, toxic fee mechanisms, liquidity traps, oracle manipulation, centralized control, and broken financial invariants. Focused on economic correctness, incentive risks, and DeFi system stability.

A research-grade framework for forecasting tokenomic gene evolution across market cycles. Analyzes historical gene frequencies, models behavioral drift, and predicts future gene expression using interpretable trend and moving-average forecasting. Designed for tokenomics research, risk analysis, and evolutionary cryptoeconomics.

Automatic optimal discretization pipeline

Leakage-safe temporal graph fraud detection and investigation workbench

A regime-switching Monte Carlo engine for simulating equity price-path distributions and quantifying tail risk.

Fruit fly introduction risk dashboard for USDA APHIS PPQ. Dual frontend: Leaflet.js geospatial dashboard + Streamlit analytics platform (Poisson GLM, Pseudo-R² 0.68). Covers 6 invasive species across air passenger & cargo pathways. USDA × CSU Hackathon 2026.

End-to-end credit risk pipeline: leakage-safe PD modeling, holdout evaluation, and full-portfolio risk band segmentation (50k loans) with stakeholder-ready visuals + Streamlit scoring app.

AI-powered financial health forecasting for retail banking segments — Flask + statistical modeling + GenAI. Built for NatWest Code for Purpose Hackathon 2026.

Flood-retention investment MVP that ranks where extra capacity can reduce downstream risk first.

Code that implements Factor Analysis of Information Risk (FAIR) in combination with MITRE ATT&CK using Baysian networks (via PyMC) to determine the frequency of successful attacks.

Quantitative model for measuring organizational security risk caused by human dependencies, decision concentration, and bus-factor effects.

Credit-default risk model: leakage-safe pipeline, calibrated probabilities, cost-based decision threshold and error analysis (scikit-learn)

Production-ready Credit Risk Decision Engine using ML, Expected Loss & FastAPI with Streamlit UI

Public-safe, runnable reconstruction of breast cancer risk-model external validation, calibration, local adaptation, and temporal validation.

Framework for regime risk modeling with FX basis, scenario shocks, and LC allocation under controls.

Severity and frequency fitting and aggregate loss distributions: transformed beta and gamma catalogs, truncation- and censoring-aware MLE, FFT and Panjer.

Real-time Fraud Detection System with Cost-Based Decision Engine (ML + FastAPI + Streamlit)

Location-level Florida hurricane catastrophe model in Python: HURDAT2-calibrated hazard + physical wind field (Holland, Vickery-Wadhera, Kaplan-DeMaria), synthetic exposure, HAZUS-anchored vulnerability, OED/YLT/ELT formats, XoL reinsurance with reinstatements, gross/net AEP/OEP curves & PMLs, historical backtesting (Andrew, Ian).