Play 50 — Financial Risk Intelligence

AI-powered financial risk platform — explainable credit risk scoring (ECOA/GDPR-compliant), three-tier real-time fraud detection (rules→ML→LLM), market sentiment analysis, regulatory reporting (Basel III, SOX), fairness testing across protected attributes, and immutable audit trails.

Architecture

Component	Azure Service	Purpose
Risk Analysis	Azure OpenAI (GPT-4o)	Credit scoring, edge-case fraud analysis
Fraud ML Model	Local (scikit-learn)	Fast statistical fraud scoring (<10ms)
Fraud Rules	Custom engine	Velocity, amount, geo-impossible checks (<1ms)
Audit Trail	Azure Cosmos DB	Immutable, SOX-compliant decision logging
Event Stream	Azure Event Hubs	Real-time transaction ingestion
Risk Engine	Azure Container Apps	Scoring API with auto-scaling
Secrets	Azure Key Vault	API keys, connection strings

📐 Full architecture details

Aspect	Play 35 (Compliance Engine)	Play 50 (Financial Risk)	Play 45 (Event AI)
Domain	General regulatory compliance	Financial services	Any event stream
Decisions	Compliance gap detection	Credit approve/decline + fraud block/allow	Anomaly detection
Regulation	GDPR, SOC 2, EU AI Act	ECOA, GDPR Art.22, Basel III, SOX	N/A
Explainability	Compliance report	Per-decision factor list (ECOA adverse)	Alert reason
Fairness	General bias testing	Protected attribute testing (4/5 rule)	N/A
Latency	Batch analysis	< 100ms real-time (fraud rules+ML)	< 500ms streaming
Audit	Compliance trail	7-year immutable audit (SOX)	Event logs

DevKit Structure


50-financial-risk-intelligence/
├── agent.md                                    # Root orchestrator with handoffs
├── .github/
│   ├── copilot-instructions.md                 # Domain knowledge (<150 lines)
│   ├── agents/
│   │   ├── builder.agent.md                    # Credit scoring + fraud + sentiment
│   │   ├── reviewer.agent.md                   # Explainability + bias + compliance
│   │   └── tuner.agent.md                      # Thresholds + fairness + cost
│   ├── prompts/
│   │   ├── deploy.prompt.md                    # Deploy risk engine
│   │   ├── test.prompt.md                      # Test scoring + fraud scenarios
│   │   ├── review.prompt.md                    # Regulatory audit
│   │   └── evaluate.prompt.md                  # Accuracy + fairness metrics
│   ├── skills/
│   │   ├── deploy-financial-risk-intelligence/ # Three-tier fraud + credit + audit
│   │   ├── evaluate-financial-risk-intelligence/ # AUC-ROC, recall, fairness, compliance
│   │   └── tune-financial-risk-intelligence/   # Thresholds, fairness, explainability
│   └── instructions/
│       └── financial-risk-intelligence-patterns.instructions.md
├── config/                                     # TuneKit
│   ├── openai.json                             # Risk model (temp=0, seed=42)
│   ├── guardrails.json                         # Fraud thresholds, fairness, audit
│   └── agents.json                             # Audit retention, regulatory rules
├── infra/                                      # Bicep IaC
│   ├── main.bicep
│   └── parameters.json
└── spec/                                       # SpecKit
    └── fai-manifest.json

Quick Start


# 1. Deploy risk engine infrastructure
/deploy
 
# 2. Test credit scoring + fraud detection scenarios
/test
 
# 3. Run regulatory compliance audit
/review
 
# 4. Measure accuracy + fairness
/evaluate

Key Metrics

Metric	Target	Description
Credit AUC-ROC	> 0.80	Score discriminative power
Fraud Recall	> 95%	True fraud detected
Fraud FPR	< 2%	Legitimate transactions blocked
Disparate Impact	> 0.80	Fair lending 4/5 rule compliance
Adverse Action Notice	100%	ECOA mandatory on declines
Fraud Detection Latency	< 100ms	Rules + ML tiers combined

Estimated Cost

Service	Dev/mo	Prod/mo	Enterprise/mo
Azure OpenAI	$60	$600	$2,200
Azure AI Search	$0	$250	$800
Cosmos DB	$5	$200	$700
Azure Event Hubs	$12	$90	$1,200
Azure Functions	$0	$200	$600
Azure Stream Analytics	$25	$150	$600
Key Vault	$1	$10	$25
Application Insights	$0	$40	$120
Total	$103	$1,540	$6,245

Estimates based on Azure retail pricing. Actual costs vary by region, usage, and enterprise agreements.

💰 Full cost breakdown

WAF Alignment

Pillar	Implementation
Responsible AI	Explainable scoring, fairness testing, ECOA adverse action notices
Security	Immutable Cosmos DB audit trail, no raw PII in logs, Key Vault
Reliability	Three-tier fraud (rules→ML→LLM), deterministic scoring (temp=0, seed=42)
Cost Optimization	LLM only for uncertain zone (<10%), local ML model, rule pre-screening
Operational Excellence	SOX audit trail, Basel III model cards, weekly fairness testing
Performance Efficiency	<1ms rules + <10ms ML = <100ms avg fraud decision

FAI Manifest

Field	Value
Play	`50-financial-risk-intelligence`
Version	`1.0.0`
Knowledge	R2-RAG-Architecture, O2-Agent-Coding, R3-Deterministic-AI, T2-Responsible-AI, T3-Production-Patterns, F2-LLM-Selection
WAF Pillars	security, reliability, responsible-ai, cost-optimization, operational-excellence, performance-efficiency
Groundedness	≥ 85%
Safety	0 violations max