dyvertexapi
The mobile app and the web app each talk to a slightly different version of the API.
Backend
Backends that can be operated, not just deployed.
APIs, databases, queues, and the boundaries between services. Split where splitting buys independent deploys or independent scaling, kept together where it only buys network calls.
- commit
- CI
- migration
- deploy
- production
- real load
The stage where architecture decisions stop being opinions.
You are probably here because
If none of these sound like your app, the free call is still the fastest way to find out what does.
data
One slow query takes the whole product down and nobody knows which one.
services
Someone split it into six services, and a single feature now touches four of them.
ops
Deploys are manual, and only one person is allowed to do them.
What we build
Scope agreed in writing before anyone writes code. If something outside it turns out to matter more, we re-scope in the open instead of quietly billing for it.
- API design
- One contract, typed, versioned, and shared by every client. Types generated from the source rather than a document that drifts a week after it is written.
- Data layer
- Schema, indexes, constraints, and migrations, with the queries that actually run checked against a realistic data volume rather than against twelve seed rows.
- Service boundaries
- A split where it buys independent deploys or independent scaling. No split where it only buys network calls, distributed transactions, and four repositories to change for one feature.
- Async work
- Queues, retries with backoff, idempotency, dead letters, and scheduled jobs that fail loudly instead of stopping quietly.
- Tests
- Integration tests against a real database rather than a mock, contract tests so a client cannot silently drift from the API, and a load check on the endpoints that will actually get traffic. AI is used to generate the input a real caller eventually sends: the out-of-range value, the concurrent write, the retry that arrives after the timeout. All of it in CI.
- Operations
- CI, environments, secrets out of the repository, migrations that run safely under load, structured logs, metrics, and alerts that mean something when they fire.
- Handoff
- Runbook, architecture notes, and the reasoning behind each boundary, so the next change is a decision rather than a guess.
NestJSHonoTypeScriptPostgresMySQLDrizzlePrismaRedisDockerAWSSentry
Price
Priced after we look.
Backend work varies too much to put one honest number on this page. The audit at $1,800 is the usual first step, its fix list lands in two working days, and the scope after it is fixed-price before anything starts.
We have shipped our own products to real users: Mira on the iOS App Store and ariex.fit in public beta. There are no rescue case studies in the portfolio yet, and we are not going to claim there are some under NDA.
Backend
Scoped on the call
APIs, data, queues, and boundaries that earn their keep.
Start with the free callQuestions we get asked first.
Anything not answered here is worth twenty minutes on a call.
Why is there no fixed price on this page?
Because a backend engagement can be one endpoint or a data model rebuild, and quoting one number for both would be a number we would have to walk back. The audit at $1,800 gives us both something real to price against within two working days, and the scope after it is fixed.
Should we use microservices?
Usually not yet. A split earns its cost when parts of the system need to deploy or scale independently, or when separate teams own them. Below that, it mostly buys you network calls, distributed transactions, and a longer path to production. We will say which side of that line you are on.
Can you work with our existing backend?
Yes, including ones we would not have built that way. The audit comes first so the plan is based on what is running rather than on what the architecture diagram claims.
Do you do database migrations on live systems?
Yes, with expand-and-contract rather than a maintenance window where we can, and with a rollback path either way. Any migration that cannot be reversed gets said out loud before it runs.
Which language and framework?
TypeScript with NestJS or Hono is what we reach for most, on Postgres. If your team is already productive in something else, keeping you there usually beats a rewrite, and we will say so.
Book the audit
Want a straight answer on this one?
Send the repo, the rejection notice, or just the idea. Twenty minutes is enough to tell you whether a rescue, a rebuild, or a clean no is the right next step.
Prefer email? [email protected]