ParasJain03/Credit-Risk-Scorecard ? reverse-engineered prompt

Build me a credit risk scorecard project that looks and feels like something a bank could use for loan decisions.

I want a simple Streamlit app where someone can enter borrower details like age, income, credit utilization, delinquencies, debt ratio, and number of credit lines, then instantly get a probability of default, a credit score, a risk category, and a clear loan decision like approved or rejected. Please also include the full model workflow in a notebook, from data cleaning and feature engineering to training a logistic regression model, handling class imbalance, calibrating the probabilities, and turning the results into a scorecard. If possible, add a section that estimates expected credit loss using the IFRS 9 style idea of PD times LGD times EAD, and show a few charts for model performance and explainability. Use the existing saved model if it helps, and make the app polished enough to demo.