Sentinel

Sentinel is a portfolio/MVP fraud-risk platform designed to demonstrate modern backend engineering, reliability patterns, and event-driven architecture in a BFSI/fintech context.

Multi-Repository Workspace Structure

In accordance with Sentinel's repository architecture (SENTINEL_Design_Decisions.md Section 27), the platform is structured into three independent logical repositories:

Sentinel/
├── sentinel-service/       # .NET 10 solution, API, workers, domain, persistence, messaging
├── sentinel-dashboard/     # Angular 21 & Material 3 Fraud Operations Console
└── sentinel-ml-service/    # Python data science, feature engineering, offline model training

Ingestion & Processing Architecture

Transaction Simulator (sentinel-service)
        ↓ HTTP POST
Sentinel Ingestion API (sentinel-service)
        ↓ ACID transaction
PostgreSQL + Outbox Table
        ↓ Outbox Publisher (sentinel-service)
RabbitMQ Exchange
        ↓
Fraud Worker (sentinel-service)
        ↓ Deterministic Rules + ML Inference
Risk Assessment Persisted
        ↓ Realtime Event
SignalR Hub
        ↓
Angular Fraud Operations Dashboard (sentinel-dashboard)

Architectural Boundaries

• sentinel-service: Core runtime containing all .NET components (Sentinel.Domain, Sentinel.Application, Sentinel.Contracts, Sentinel.Infrastructure, Sentinel.Api, Sentinel.FraudWorker, Sentinel.OutboxPublisher, Sentinel.Simulator).
• sentinel-dashboard: Angular application representing the fraud operations console. It communicates exclusively via HTTP REST APIs and SignalR realtime events; it contains no C# dependencies and does not submit transactions.
• sentinel-ml-service: Independent Python environment reserved for dataset analysis and model training (deferred). Runtime inference is executed natively in .NET within the backend.