RestaurantOS
White-label restaurant ERP on 15 microservices, with tenant isolation that fails closed.
Technical lead, team of four · Self-employed · 2026 – now
- 15
- Spring Boot microservices
- 100%
- policy-test coverage in CI
How it is built, step by step
- Step 1
One door
Every request enters through Spring Cloud Gateway with RS256 JWT and TOTP two-factor.
- Step 2
Policy first
OPA/Rego decides before any service runs, and it fails closed: no matching policy, no access. 100% policy-test coverage in CI.
- Step 3
Fifteen services
Point of sale, kitchen display, inventory, finance, purchasing, HR and payroll, reporting, and more, each its own Spring Boot service.
- Step 4
The database enforces tenancy too
PostgreSQL row-level security on every tenant table, so even a bug in a service cannot leak across tenants.
- Step 5
Events, cache, analytics
RabbitMQ with dead-letter queues, Redis, and ClickHouse for reporting.
- Step 6
Shipped
GitHub Actions delivers to Kubernetes (k3s).
In detail
I architected the system and lead the team: point of sale, kitchen display, inventory, finance, purchasing, HR and payroll, and reporting, on infrastructure designed so a tenant can never see another tenant.
RestaurantOS is a white-label ERP for restaurants. It runs as 15 Java/Spring Boot microservices behind Spring Cloud Gateway, with a Next.js front end, RabbitMQ with dead-letter queues, Redis and ClickHouse for reporting.
Tenant isolation is enforced in the database and in policy: PostgreSQL row-level security on every tenant table, and fail-closed OPA/Rego authorisation with 100% policy-test coverage enforced in CI. Authentication is RS256 JWT with TOTP two-factor.
Delivery is GitHub Actions to Kubernetes (k3s). As technical lead I own the architecture, the review bar and the security model for a team of four.