nexvra services --verbose

What we build, and what you get.

Three services, each scoped in writing and priced before it starts. Every engagement ships with the parts teams usually defer: tests, observability, runbooks, and documentation your next engineer can follow.

run_estimator ./pricing
01 / platform-api  ·  typically 6 – 16 weeks

Platform & API Engineering

Most products don't fail on features — they fail the first time load doubles. We build the layer underneath: well-shaped APIs, idempotent jobs, sane queueing, and observability you can actually debug at 3am.

We work in your stack rather than imposing ours. TypeScript, Go, Python, Postgres, Redis — chosen because your team can maintain it after we leave, not because it's fashionable this quarter.

deliverables
  • Architecture decision records for every significant choice
  • Versioned REST or GraphQL API with generated docs and typed clients
  • Background job and queue infrastructure with retry semantics
  • Structured logging, tracing, and alerting wired to real thresholds
  • Load-test results against agreed targets, not vibes
  • Runbooks for the five failures most likely to wake someone up
TypeScriptGoPythonPostgreSQLRedisgRPCKafkaTerraformOpenTelemetry
typical_shape
  • Fixed scope, fixed price, milestone billing
  • Two-week cycles ending in a deployed environment
  • Shared channel with the engineers writing the code
  • Architecture decisions recorded with trade-offs
  • Full source and infrastructure handover
  • 90 days of post-launch support
discuss_this
02 / data-ai  ·  typically 4 – 12 weeks

Data & AI Infrastructure

An LLM demo takes an afternoon. Making it reliable enough to put in front of customers takes engineering: retrieval that returns the right chunk, evaluation that catches regressions, and guardrails that fail closed.

Every engagement starts with an honest assessment of whether the model can hit the bar your use case needs. When it can't, we tell you before you've spent the budget rather than after.

deliverables
  • Feasibility assessment with a documented go / no-go recommendation
  • Ingestion and transformation pipelines with data-quality checks
  • Retrieval-augmented generation over your own corpus
  • Evaluation harness with regression suite and scored baselines
  • Prompt and model versioning, plus rollback capability
  • Token and inference cost model at your projected volume
Claude APIOpenAIpgvectorLangGraphdbtAirflowPythonDuckDB
typical_shape
  • Fixed scope, fixed price, milestone billing
  • Two-week cycles ending in a deployed environment
  • Shared channel with the engineers writing the code
  • Architecture decisions recorded with trade-offs
  • Full source and infrastructure handover
  • 90 days of post-launch support
discuss_this
03 / modernise  ·  typically 2 – 10 weeks

Systems Modernisation & Rescue

Legacy systems are rarely as broken as they feel — they are usually undocumented, untested, and frightening to touch. Those are different problems with different, cheaper fixes.

We start read-only: map the system, instrument it, and find where the risk actually concentrates. Only then do we recommend what to change. A full rewrite is the answer roughly one time in ten, and we will say so when it isn't.

deliverables
  • Dependency and data-flow map of the existing system
  • Prioritised risk register with severity and blast radius
  • Characterisation tests around the parts you're afraid to touch
  • Incremental migration plan with reversible steps
  • CI pipeline and environment parity where none existed
  • Written handover so your team owns it, not us
Legacy PHPJavaRails.NETDockerGitHub ActionsPostgresDatadog
typical_shape
  • Fixed scope, fixed price, milestone billing
  • Two-week cycles ending in a deployed environment
  • Shared channel with the engineers writing the code
  • Architecture decisions recorded with trade-offs
  • Full source and infrastructure handover
  • 90 days of post-launch support
discuss_this
nexvra scope --exclusions

What we decline.

Publishing this saves everyone a discovery call.

Work we turn down

  • Pure design or branding with no engineering attached
  • Marketing sites and brochureware
  • Staff augmentation with no say in the architecture
  • Projects won by undercutting rather than by scope
  • Anything covered by our Acceptable Use Policy

What we say early

  • If an existing product solves it, we name the product
  • If the budget cannot carry the scope, you hear it in call one
  • If a model cannot hit your reliability bar, we will not ship it
  • If a rewrite is not warranted, we argue against our own upside

Describe the problem, not the solution.

Tell us what is breaking, what is slow, or what will not scale. We will tell you what it takes to fix — including when the answer is that you should not build it.

start_project run_estimator