Reproducible engine that turns Australian bank and regulator disclosures (Pillar 3, APRA, RBA, S&P) into governed, auditable PD / LGD / EL model inputs — base and stressed — with a full audit trail and a 595-test suite.
End-to-end Credit Risk & Expected Loss Analytics Platform | PD Modelling, XGBoost, SHAP, Streamlit | LendingClub 890k loans
Streamlit app that computes per-loan Expected Loss, Lifetime ECL, and Risk Rating from EAD/PD/LGD/WAL Excel portfolios
Implements the Basel III credit risk framework (PD, LGD, EAD) using Logistic & Linear Regression on Lending Club loan data (2007–2014)
Governed PostgreSQL credit decisioning framework for synthetic applications, pre-production strategy simulation, matched comparison, counteroffer governance, Expected Loss tradeoffs, and Power BI release validation—without PII.
This model estimates the 12-month Probability of Default (PD) for prime residential mortgage customers in the United Kingdom, aligned with the IFRS 9 impairment framework and calibrated to an adverse macroeconomic scenario. Version 1 (v1) is developed using gradient-boosted decision trees (GBDT)
End-to-end bank credit risk scorecard: WoE/IV logistic regression (Gini 0.71) with fair-lending design, LightGBM + SHAP challenger, cut-off and expected-loss analysis, and an AI credit-memo assistant (RAG, Claude API) that cites bank policy, all in a Streamlit app with PDF credit reports.
Deterministic expected-loss change attribution and audit evidence for credit portfolios
End-to-end credit risk analytics project using calibrated gradient boosting, economic threshold optimization, expected loss, explainability, and fairness analysis.
End-to-end credit risk engine computing Expected Loss (PD × LGD × EAD) on the LendingClub portfolio.
Actuarial insurance risk scoring using frequency–severity modelling to estimate expected loss for underwriting and pricing.
Interactive collections intelligence dashboard using Streamlit, SQL, and Expected Loss analytics for portfolio risk analysis and collections prioritization.
Proyecto de Titulación: Cálculo de Pérdidas Esperadas basado en 3 modelos de Credit Scoring para una institución financiera del Ecuador
Credit risk PD modeling, expected loss, policy simulation, explainability and Streamlit dashboard.
Probability-of-default models and expected credit loss (PD × LGD × EAD) on 30,000 consumer credit accounts, with calibration, stress scenarios, SHAP explainability, model validation and interactive risk dashboards.
End-to-end SME credit analytics platform built with R, SQL, and Shiny for loan portfolio analysis, credit scoring, risk monitoring, expected loss estimation, and management dashboards.
End-to-end credit risk pipeline for PD, LGD, EAD, expected loss, IFRS 9-style staging, and stress testing on LendingClub loan data.
End-to-end credit risk modeling: PD/LGD/EL on 1.35M Lending Club loans. Test AUC 0.719, 32 bps portfolio error. Includes drift monitoring, MRM documentation, and live dashboard.
Personal-loan credit-risk & unit-economics platform: vintage delinquency curves, expected-loss modeling, and an interactive approval-threshold scenario simulator. Built with dbt, DuckDB, Python, and Streamlit.
To associate your repository with the expected-loss topic, visit your repo's landing page and select "manage topics."