Skip to content
Now taking Q3 engagements

Ship thehard parts.

Codeaex is a senior engineering studio. We build the distributed systems, data platforms, and AI infrastructure that other teams stall on — and we stay on call for them afterwards.

systems in production
140+

systems in production

fleet-wide uptime
99.98%

fleet-wide uptime

average engineer tenure
11 yrs

average engineer tenure

// Blue/green rollout with automatic rollback
export async function release(svc: Service) {
const next = await svc.provision({ replicas: 12 })
await next.warm()
if (await healthy(next, { p99: 80 })) {
return svc.cutover(next)
}
await svc.rollback()
throw new ReleaseError("p99 regression")
}
typecheck passing
zsh — codeaex
$

Built with

Technologies we work in: TypeScript, Go, Rust, Python, React, Next.js, Node.js, PostgreSQL, Kafka, Redis, Kubernetes, Terraform, AWS, gRPC, ClickHouse, OpenTelemetry
Reference architecture

This is what we hand over.

A composite of the systems we build: typed at every boundary, observable end to end, and deployable by your team without us in the room.

system.topologylive traffic
EDGEAPPLICATIONDATAOPERATIONSClientsWeb · Mobile · Partner APIEdge / CDNCache · WAF · TLSAPI GatewayAuthN/Z · Rate limitServicesGo · gRPC · 47 deploysWorkersAsync jobs · retriesOLTPPostgreSQL · replicasEvent BusKafka · exactly-onceWarehouseClickHouse · dbtTelemetryOpenTelemetry · tracesCI / CDIaC · progressive delivery

Scroll to pan the diagram

Typed at every boundary

Schemas are versioned and generated into clients. No untyped JSON crossing a service line, and no silent contract drift between teams.

Observable end to end

A single trace carries a tenant id from the edge through the queue to the query plan. When something is slow, you know which hop owns it.

Reproducible from zero

Every environment can be rebuilt from a clean cloud account with Terraform. Nothing important exists only in someone's console history.

Reversible by default

Progressive delivery, feature flags, and automatic rollback on SLO breach. A bad release is an inconvenience, not an incident.

Fleet telemetry

Numbers we are held to.

Every system we run reports into the same dashboard. Rolling 90-day figures across the fleet.

p99 latency

-18% QoQ

0ms

Uptime

SLO 99.9%

0.00%

Deploys / week

+37% QoQ

0

Change failure rate

elite band

0.0%

Process

Four stages, no surprises.

The same sequence whether the engagement runs eight weeks or two years. Predictability is a feature.

  1. 01

    Technical discovery

    Week 1

    We read the code, trace the hot paths, and interview the people who get paged. You get a written assessment of what is actually in the way — including the parts you did not ask about.

    You receive: Assessment + risk register

  2. 02

    Architecture & plan

    Weeks 2–3

    Target architecture, migration sequence, delivery plan, and a fixed commercial shape. Enough detail to decide with your eyes open, and cheap enough to walk away from.

    You receive: ADRs + sequenced roadmap

  3. 03

    Build in increments

    Weeks 4–16

    Two-week increments, each deployed behind a flag and demonstrable in your environment. You see running software, not slides about running software.

    You receive: Deployed increments + CI

  4. 04

    Handover & on-call

    Ongoing

    Runbooks, architecture decision records, and a rota sized to how much of it you want to own. We answer the pager until you would rather we didn't.

    You receive: Runbooks + SLOs + rota

References

What the people who hired us say.

Every reference below will take a call. Ask us and we will make the introduction.

Three vendors told us the migration would take three years and break everything. Codeaex did it in eleven months with zero unplanned outages. I have never worked with a team so consistently, boringly right.

Eleanor Vance

CTO · Meridian Bank

They argued with us — politely, with benchmarks — about half of what we asked for. The system is better for every one of those arguments.

Dr. Samuel Okafor

Clinical Director · Halcyon Health

Two years after handover our own team owns it end to end, and the on-call load went down, not up. That is the whole review.

Priya Raman

VP Engineering · Ardent Logistics

Engagements

Tell us what is breaking.

Send the problem, not a polished brief. A principal engineer reads every enquiry and replies within two working days with a real technical opinion.

  • Reply from an engineer, not a salesperson
  • Fixed scope and price before any build starts
  • Your code, your cloud, your repo — from day one