Skip to content

README.md

Latest commit

 

History

History
118 lines (91 loc) · 5.76 KB

README.md

File metadata and controls

118 lines (91 loc) · 5.76 KB
Raw

Credit Risk Assessment

Overview

The Credit Risk Assessment use case evaluates borrower creditworthiness through coordinated financial statement analysis, quantitative risk scoring, and portfolio impact assessment. It produces credit ratings, probability-of-default estimates, and portfolio concentration analysis to support lending decisions by credit committees.

Business Value

  • Comprehensive credit views -- three specialist agents cover financial health, risk quantification, and portfolio impact in a single request
  • Faster underwriting -- parallel agent execution compresses multi-day manual analysis into minutes
  • Consistent methodology -- standardized risk scoring with PD/LGD estimation and credit rating assignment (AAA through D)
  • Portfolio awareness -- every borrower assessment includes concentration risk and diversification impact analysis
  • Decision-ready output -- structured response with credit rating, score, and actionable recommendations for credit committees

Architecture

graph TB
    Request["Client Request"] --> Runtime["AgentCore Runtime"]
    Runtime --> Orchestrator["Orchestrator"]
    Orchestrator --> Financial["Financial Analyst<br/><small>Statement analysis & ratio computation</small>"]
    Orchestrator --> Risk["Risk Scorer<br/><small>Credit scores & PD/LGD estimation</small>"]
    Orchestrator --> Portfolio["Portfolio Analyst<br/><small>Concentration & diversification metrics</small>"]
    Financial --> Bedrock["Amazon Bedrock<br/>(Claude)"]
    Risk --> Bedrock
    Portfolio --> Bedrock
    Financial --> S3["S3 Sample Data"]
    Risk --> S3
    Portfolio --> S3
    Financial --> Synthesis["Result Synthesis"]
    Risk --> Synthesis
    Portfolio --> Synthesis
    Synthesis --> Response["Response"]
Loading

Directory Structure

use_cases/credit_risk/
├── README.md
└── src/
    └── strands/
        ├── __init__.py
        ├── config.py          # CreditRiskSettings
        ├── models.py          # Pydantic request/response models
        ├── orchestrator.py    # CreditRiskOrchestrator + run_credit_risk()
        └── agents/
            ├── __init__.py
            ├── financial_analyst.py
            ├── risk_scorer.py
            └── portfolio_analyst.py

Agentic Design

The orchestrator uses a parallel fan-out pattern. In full mode, all three agents execute concurrently via asyncio.gather. Individual modes (financial_analysis, risk_scoring, portfolio_analysis) invoke a single agent. The orchestrator synthesizes combined findings into a structured JSON credit assessment with scoring, rating, and portfolio impact.

Agents

Agent Role Data Used Output
Financial Analyst Analyzes income statements, balance sheets, and cash flows; computes debt-to-equity, current ratio, interest coverage ratios Borrower profile via s3_retriever_tool Revenue/profitability trends, key ratios with benchmarks, cash flow adequacy, financial health summary
Risk Scorer Computes credit risk scores, estimates probability of default (PD) and loss given default (LGD), assigns credit ratings Borrower profile via s3_retriever_tool Risk score (0-100), risk level, credit rating (AAA-D), PD/LGD estimates, risk factors and mitigants
Portfolio Analyst Evaluates portfolio concentration by sector/geography/counterparty, calculates diversification metrics and risk-adjusted returns Borrower profile via s3_retriever_tool Concentration change, diversification score (0-1), sector exposure, risk-adjusted return, portfolio notes

Data and Tools

  • Tool: s3_retriever_tool -- retrieves borrower profiles and financial data from S3
  • S3 data prefix: samples/credit_risk/
  • Model: Claude Sonnet (via Amazon Bedrock), temperature 0.1, max 8192 tokens
  • Config thresholds: risk_threshold_high=75, risk_threshold_critical=90, max_portfolio_concentration=0.25

Request / Response

Request -- AssessmentRequest:

Field Type Description
customer_id str Borrower identifier (e.g., BORROW001)
assessment_type AssessmentType full, financial_analysis, risk_scoring, portfolio_analysis
additional_context str | None Optional context

Response -- AssessmentResponse:

Field Type Description
customer_id str Borrower identifier
assessment_id str Unique assessment UUID
timestamp datetime Assessment timestamp
credit_risk_score CreditRiskScore | None Score (0-100), level, rating (AAA-D), PD, LGD, factors, recommendations
portfolio_impact PortfolioImpact | None Concentration change, diversification score, sector exposure, risk-adjusted return
summary str Executive summary
raw_analysis dict Raw agent output

Quick Start

# Deploy to AgentCore
USE_CASE_ID=credit_risk ./scripts/deploy/full/deploy_agentcore.sh

# Test the deployment
./scripts/use_cases/credit_risk/test/test_agentcore.sh

Sample Data

Located at data/samples/credit_risk/

Borrower ID Industry Description
BORROW001 Manufacturing Acme Manufacturing Corp -- $50M revenue, D/E 1.29, current ratio 1.85, interest coverage 4.2x, requesting $20M term loan for expansion with $33M in collateral (real estate + equipment)

Related Documentation