Problem
Reliability work is hard to judge from screenshots alone. PulseBoard needed to expose the system boundaries, failure behavior, and operational controls that make a small SaaS credible.
My role
I designed and implemented the public product surface, API contract, persistence model, worker flow, deployment plan, and verification suite.
Constraints
The system had to remain inexpensive, locally reproducible, explicit about authentication boundaries, and deployable without committing secrets.
Key decisions
Hono keeps the API small, PostgreSQL owns durable state, Redis and BullMQ isolate asynchronous checks, and Nginx preserves a stable public namespace under /demo/.
Architecture
Requests cross an API-key boundary before reaching workspace-scoped resources. Jobs enter a queue, workers execute checks, and health endpoints expose liveness and dependency readiness separately.
Engineering evidence
The live console, customer view, OpenAPI document, health probes, tests, ADR, deployment runbook, and public-surface verifier form the evidence set.
Result
The result is an inspectable portfolio system: visitors can move from product behavior to API and operational proof without relying on invented usage metrics.
Limits & next steps
This remains a portfolio-scale deployment. Load testing, multi-region failover, and production alert delivery are intentionally outside the current evidence boundary.