Play 63
Fraud Detection Agent
High✅ Ready
Real-time transaction fraud detection with sub-100ms latency and explainable decisions.
Agentic fraud detection system that analyzes transactions in real time using behavioral anomaly scoring, velocity checks, device fingerprinting, and network graph analysis. Azure Event Hubs ingests transaction streams, Stream Analytics applies windowed pattern detection, and GPT generates human-readable explanations for every flagged decision — meeting regulatory compliance requirements. Achieves sub-100ms p99 latency with full audit trails in Cosmos DB.
Architecture Pattern
Real-time streaming: multi-signal scoring, deterministic rules, explainable decisions
Azure Services
Azure OpenAIEvent HubsStream AnalyticsCosmos DBAzure FunctionsAzure Monitor
DevKit (.github Agentic OS)
- agent.md — root orchestrator with builder→reviewer→tuner handoffs
- 3 agents — Fraud Builder (gpt-4o), Reviewer (gpt-4o-mini), Tuner (gpt-4o-mini)
- 3 skills — deploy (211 lines), evaluate (121 lines), tune (182 lines)
- 4 prompts — /deploy, /test, /review, /evaluate with agent routing
- .vscode/mcp.json — FrootAI MCP with OpenAI key input + envFile
TuneKit (AI Config)
- config/openai.json — gpt-4o-mini for real-time, gpt-4o for investigation
- config/guardrails.json — latency budget, explainability rules
- evaluation/eval.py — Precision >95%, Latency p99 <100ms
Tuning Parameters
Risk score thresholdVelocity window secondsAnomaly sensitivityExplainability depthLatency budget ms
Estimated Cost
Dev/Test
$100–250/mo
Production
$3K–12K/mo