Commercial Portfolio Monitoring & Early Risk Analytics Project

This repository is the post-origination monitoring and impairment reporting layer in the commercial credit-risk stack. It uses loan-level expected loss data, prior-period snapshots, and optional pricing or capital inputs to produce facility-level ECL outputs, stage summaries, transition matrices, early-warning views, and concentration reports. The project is positioned as a portfolio monitoring layer that is relevant to both bank-style impairment review and non-bank portfolio performance management.

What this repo is

This project demonstrates how a commercial portfolio can be monitored after origination using staging, lifetime PD, scenario-weighted ECL, migration, and concentration analytics. It is designed as a portfolio project, so the workflow is transparent, the assumptions are clearly disclosed, and the outputs are shaped for recruiter-friendly review across both institutional and non-bank lending contexts.

Where it sits in the stack

Upstream inputs:

• expected-loss-engine-commercial
• optional facility-level capital context from RWA-capital-commercial
• optional pricing context from RAROC-pricing-and-return-hurdle
• prior-period snapshots and reviewer inputs staged under data/

Downstream consumers:

• portfolio monitoring packs and early-warning review
• impairment and stage-movement summaries
• employer-ready disclosure and management-reporting examples

How this is used in practice

This project can be applied in:

Bank / Institutional context

• Portfolio risk monitoring, staging, and impairment-style reporting
• Early-warning and migration analysis for structured risk review
• Concentration, watchlist, and management reporting support

Non-bank / Fintech context

• Early risk monitoring and portfolio performance tracking after origination
• Roll-rate, arrears-style, and warning-signal review for collections or servicing teams
• Management reporting on customer cohorts, segments, and emerging risk pockets

Key outputs

• data/output/facility_ecl.csv
• data/output/ecl_summary_by_stage.csv
• data/output/ecl_summary_by_segment.csv
• data/output/concentration_report.csv
• data/output/transition_matrix_grade.csv
• data/output/transition_matrix_stage.csv
• data/output/early_warning_summary.csv
• data/output/aps330_stage_movement.csv
• data/output/aps330_credit_quality.csv

Repo structure

• data/: tracked folder guide plus runtime-created input/, manual/, processed/, and output/ subfolders used during local runs
• src/: reusable staging, lifetime PD, ECL, migration, and monitoring modules
• scripts/: wrapper scripts for pipeline execution
• docs/: methodology and disclosure notes
• notebooks/: reviewer-facing notebook index and walkthrough placeholders
• tests/: validation and regression checks

How to run

python -m src.pipeline --refresh-demo-inputs

Or:

python scripts/run_pipeline.py --refresh-demo-inputs

Run validation tests:

pytest

Limitations / Demo-Only Note

• All monitoring inputs are synthetic or simulated for demonstration purposes.
• The repo shows a practical monitoring and impairment workflow, not a production accounting, disclosure, or regulatory reporting platform.
• Scenario weights, staging thresholds, and early-warning rules are simplified so the logic remains easy to inspect.