Creating a Contemporary Lending Risk Management System Using Python
summary
A complete lending risk stack built on the PyData ecosystem uses rpy2 for R forecasting and Spyre for a lightweight web UI, running millions of models nightly across 200,000 suppliers to generate per-supplier 90‑day forecasts, XGBoost classifiers for AR discontinuation, PACER-based bankruptcy matching, invoice congruency checks and voice NLP fraud flags; portfolio rules set credit limits and rates to target 6.25% in 40 days and monitoring triggers kept the fund with no defaults in 1.5 years.
AI-written summary · about
chapters (15)
- 0:00Modern lending risk systems
- 3:13Water for Commerce Fund
- 6:26Open source tools
- 7:09Concentration risk mitigation
- 8:56Forecasting AR default risk
- 13:10Seasonality in cash flow
- 14:14Predicting AR discontinuation
- 18:07Bankruptcy prediction
- 24:56Fraud risk detection
- 27:04Portfolio management with WFC score
- 30:58Portfolio health monitoring
- 33:04Luigi and Dask
- 34:30Web app demo
- 36:37Key takeaways
- 39:21Q&A
related talks
39mMachine Learning in Financial Credit Risk Assessment
31mBuilding Data-Driven Client Relationship Management in Banking with Python
36mA Data Science Approach to Systemic Risk
33mAssurance Scoring Using Machine Learning and Analytics to Reduce Risk in the Public Sector
36mDigital, data-driven financial advisory
02h 00mData Science Driven Business Insights with Python and PySp
42mRobust Automated Forecasting in Python and R
01h 30mAI in Finance From Insight to Action