Platform & API Engineering
Backends that survive contact with real traffic. APIs, services, queues, and the boring reliability work most teams skip.
TypeScriptGoPythonPostgreSQL
nexvra --version // engineering-led, 2026
We build the layer
We build the layer
your product stands on
Platform and API engineering, data and AI infrastructure, and rescue work on systems nobody wants to touch. Written scope, fixed price, and the reliability work included rather than deferred to a phase two that never comes.
$ nexvra map --repo ./your-system
reading 412 files · 3 services · 1 database
✔ dependency graph resolved
⚠ 6 risks found — 2 high, 4 medium
✔ written map delivered
$ nexvra scope --fixed-price
every service, endpoint and migration itemised
out-of-scope listed explicitly
→ price locked. this is the number you pay.
nexvra services --list
Three problems we solve properly.
We are not a general agency. We take on work where the hard part is engineering, and we decline the rest rather than learn on your budget.
Data & AI Infrastructure
Pipelines, retrieval systems, and model integration built to be measured — with evals and cost ceilings from day one.
Claude APIOpenAIpgvectorLangGraph
Systems Modernisation & Rescue
Inherited a codebase nobody understands? We map it, stabilise it, and give you a sequenced plan instead of a rewrite pitch.
Legacy PHPJavaRails.NET
nexvra cat principles.toml
Opinions we hold, in writing.
Most studios say "we care about quality". Here is what that actually means in practice, stated specifically enough that you can hold us to it.
no_rewrites_by_defaultA rewrite is the answer about one time in ten. We look for the cheaper intervention first.
boring_technologyProven tools your team can hire for, not whatever trended on Hacker News this month.
written_over_spokenScope, architecture decisions, and change orders are documents you keep. Calls are not artefacts.
reversible_stepsEvery migration ships behind a flag with a rollback path. Big-bang cutovers are how weekends die.
measure_then_optimiseNo performance work without a profile. No AI feature without an eval. No claim without a number.
leaving_is_easyYour repos, your cloud, your docs, from day one. Lock-in is not a retention strategy.
nexvra process --steps
Five phases, same every time.
Whether it is a five-week rescue or a twelve-month platform build, the sequence does not change.
01 · map
Map the terrain 3 – 5 days
Before estimating anything we read the code, the data model, and the incident history. You get a written map of the system and an honest list of what worries us. If the honest answer is that you do not need us, that is what the document says.
02 · scope
Scope and price 2 – 4 days
A scope document listing every service, endpoint, and migration in the engagement — and explicitly what is out. Architecture decisions are recorded with their trade-offs. The price attached to that document is the price you pay.
03 · build
Build in the open ongoing
Two-week cycles, every one ending in a deployed environment you can hit. You have commit access, board access, and a channel with the engineers. No status theatre, no progress percentages invented on a Friday.
04 · harden
Harden before launch 1 – 2 weeks
Load tests against the numbers in your scope document. A security pass. Failure injection on the paths that matter. Runbooks written for the failures we consider most likely. Launch is a scheduled, rehearsed event.
05 · handover
Hand over completely 30 – 90 days
Your repos, your cloud accounts, your documentation. A recorded walkthrough for whoever inherits it. Our access is revoked at handover unless you retain us. Leaving should be easy — that is what makes staying a choice.
nexvra pricing --engagements
Three ways to work with us.
Published starting numbers. The scope document confirms the final figure before any work begins — and before you owe anything.
--audit
Technical Audit
An independent read on a codebase, architecture, or vendor proposal.
$3,500flat
5 business days · fixed fee · fully credited against a later build
- Full codebase and architecture review
- Prioritised risk register with severity ratings
- Security and dependency vulnerability scan
- Scalability assessment against your growth targets
- Written report plus a 90-minute walkthrough call
- Fee credited in full if you proceed to a build
most engagements
--build
Product Build
A scoped system delivered end to end, with the reliability work included rather than deferred.
$16,500from, fixed scope
Typically 6 – 16 weeks · 40 / 30 / 30 milestone billing
- Everything in Technical Audit
- Architecture decision records for every major choice
- Full implementation with automated test coverage
- CI/CD, staging parity, and infrastructure as code
- Observability: logging, tracing, alerting, dashboards
- Load testing against agreed performance targets
--embed
Embedded Squad
Senior engineers inside your team, on your board, in your standups.
$9,800/ month
Minimum 3 months · 30 days' notice to cancel
- Two senior engineers, named and consistent
- Working in your repos, your board, your rituals
- Architecture guidance and code review for your team
- On-call participation available
- Monthly engineering health report
- Scale the squad up or down each quarter
Not sure where you land? run the estimator — it returns a real range in about a minute.
nexvra faq --all
Questions, answered directly.
How is a fixed price possible on complex engineering?
Because we never quote before the mapping phase. Reading the system first is what converts unknowns into line items. The scope document lists every service and migration we are building, and equally what we are not.
When something genuinely unforeseeable appears — an undocumented dependency, a vendor API that behaves differently than its docs — we price it as a change order and you decide. We do not absorb it silently and we do not surprise you with it later.
Who owns the code?
You do, in full, on final payment. That includes source, infrastructure definitions, and documentation, and it is written into our Terms of Service rather than promised verbally.
We retain only our own pre-existing internal tooling, and you get a perpetual royalty-free licence to keep using it inside your product.
Will you work in our existing stack?
Yes, and we prefer to. The best stack is usually the one your team can already maintain. We will flag a technology choice that we think is a genuine liability, with reasoning, but it stays your call.
The exception is where a choice makes the agreed reliability targets unreachable. In that case we say so in the scope document before you sign it, not after.
What if the audit says we should not build?
Then it says that. Roughly one audit in six concludes that an off-the-shelf product, a configuration change, or simply doing nothing beats a custom build.
You keep the report either way, and the audit fee is credited against a build only if you decide to proceed. We would rather lose the project than take budget for work that should not happen.
How does the estimator relate to the real price?
It is a calibrated range built from our actual delivery data, not a lead-capture toy. For a well-defined project the final scoped price usually lands inside it.
It is still an estimate produced without seeing your system. The mapping phase is what turns it into a fixed number, and that number can move in either direction.
Do you take on-call or production responsibility?
With an incident retainer, yes — an agreed response window, a rotation, and escalation paths documented before anything goes wrong.
Without one we are not in your on-call path, but every build ships with runbooks so your team is not debugging blind.
Ready to scope it properly?
Send us the shape of the problem. You get a written map, a fixed price, and an honest answer about whether it should be built at all — usually within two business days, at no cost.