How we built Kinaura's backend on a near-zero infra budget
Kinaura is a live consumer product we ship and run as OpsPal's own platform. The constraint was real: founder-stage budget, production reliability, and room to grow without replatforming in six months.
Stack choices
- Supabase — Postgres, auth, row-level security, and realtime where we needed it. One control plane for data + identity.
- Railway — API workers and background jobs with predictable deploys and staging parity.
- Edge-friendly clients — React Native / web clients talking to typed APIs, not direct DB access from the browser.
What kept cost down
- Managed Postgres instead of self-hosted k8s for v1.
- RLS policies defined up front—security without a custom auth monolith.
- Observability from day one (structured logs + error tracking) so we did not pay in outage time later.
- Migrations in-repo; no manual schema drift between staging and prod.
What we would not skip
"Cheap" does not mean skipping acceptance criteria. We still wrote API contracts, load-tested hot paths, and documented runbooks for on-call—even when the on-call rotation was two people.
Outcome
Production backend powering Kinaura—see the portfolio write-up for scope and stack summary.
Building something similar? App development is scoped as fixed phases; see our Kinaura platform case for the production stack in context.