SENTINEL

## context

Tracking entities — vessels, aircraft, vehicles, persons, facilities — in real time requires ingesting high-throughput feeds, answering spatial queries fast, and pushing live updates to analysts. SENTINEL is a ground-up build of that pipeline: an Nx monorepo with NestJS microservices, a Go ingest service, Python analytics, and an Angular 19 frontend on CesiumJS.

## approach

• Event-driven core: all entity state changes flow through Kafka topics, each service maintaining its own read model
• Redis GEOSEARCH geospatial cache for sub-millisecond bounding-box queries, with PostgreSQL/PostGIS fallback and automatic cache re-warming
• TimescaleDB hypertables for track history, OpenSearch for full-text and geospatial search
• Keycloak OIDC with role- and classification-based access control (UNCLASSIFIED → TOP_SECRET)
• Protobuf contracts shared across TypeScript, Go, and Python via generated libraries

## outcome

A fully containerized platform — Docker Compose for local dev, Kubernetes manifests for deployment — where live entity positions stream to a 3D map over WebSocket, with link-analysis graphs, geofence alerting, and pattern analytics running as independent services.