Transportation: a developer platform and roughly 3x more deploys a day
A team of 6 has run platform engineering, DevOps and SRE for a US transportation services company since December 2025.
Before and after, from the case
~3x
more production deployments per day since we started
~USD 5,000
a month less in observability spend
18 → 135
dashboards, with OpenTelemetry
4 → 31
repositories on build-once promotion
Drawn from the case figures. The client is not named and its product is not shown.
The engagement
- The client
- A US transportation services company
- Status
- Ongoing, phase 1 delivered
- Team
- 6 people: lead and architect, 2 platform engineers, DevOps engineer, SRE, scrum master
What changed for the engineering organization since December 2025?
Figures are organization-wide, from the client's engineering-metrics dataset built from its pull requests. The team grew and adopted AI tooling in the same period, so they describe the organization we support, not a result we claim alone.
Production deployments per day are roughly 3 times what they were when the engagement began in December 2025. Dashboards went from 18 to 135 on OpenTelemetry, and observability spend is about USD 5,000 a month lower. Build-once image promotion went from 4 to 31 repositories and shared CI from 35 to 76. Product services hold about 90% median test coverage, and the quality gate passes on 89% of projects.
~3xmore production deployments per day since the engagement began (organization-wide)
~USD 5,000a month less in observability spend
~90%median test coverage on product services
Also on record
- Dashboards from 18 to 135, on OpenTelemetry
- Build-once image promotion from 4 to 31+ repositories
- Shared CI from 35 to 76+ repositories
- Quality gate passing on 89% of projects
What we did, by area
Areas of work
The client
A growing cloud-native estate: 150+ repositories on multi-cluster Kubernetes
The client is a US transportation services company with more than 150 active repositories on multi-cluster Kubernetes on a major cloud. We publish it without its name, as agreed.
Before the engagement, releases were trunk-based with code freezes across about 15 repositories, monitoring was per cluster with 18 dashboards and no OpenTelemetry, and there was no developer portal, catalog or golden path.
Internal developer platform
A Backstage portal with 11 self-service templates
The portal has been live since spring 2026 in a highly available setup. It offers 11 self-service templates and golden-path skeletons for five stacks, including .NET, Python and Node.js.
- About 66 repositories and 73 components in the catalog.
- More than 35 backend plugins.
- Self-service infrastructure with Crossplane on all three clusters: 54 managed resources through GitOps and 15 service databases on a shared PostgreSQL per environment.
Release engineering and GitOps
Build once, promote the same image, and manage every cluster from Git
A release-branch model with automated cuts, back-merges and drift checks replaced code freezes. 17 repositories run on it, with 34 release candidates and 10 production releases since September 2026.
- Build-once image promotion: from 4 to more than 31 repositories.
- Shared CI library: from 35 to more than 76 repositories, with 9 new reusable workflows.
- Argo CD: from 46 to 101 applications and from 8 to 29 application sets, plus a management cluster added in April 2026.
Observability
From 18 per-cluster dashboards to 135 on one open-source stack
OpenTelemetry collectors run on every cluster and send to central Mimir, Loki, Tempo and Pyroscope, read in Grafana with single sign-on. Real-user monitoring, OpenCost and engineering-data exporters complete it.
There are 135 dashboards today, 34 of them per service or for DORA metrics. Moving metrics and logs off a commercial APM cut observability spend by about USD 5,000 a month.
DevSecOps and engineering metrics
A shared security workflow and a DORA metrics programme
Since February 2026 a shared security workflow runs dependency, secret, CodeQL and container scans. CI supply-chain hardening is in progress.
The DORA programme is live: deployments per day went from about 4.4 to about 12 to 16 across the organization.
Team
Six people inside the client's engineering organization
A lead and architect, 2 platform engineers, a DevOps engineer, an SRE and a scrum master, since December 2025. The team delivered 464 platform and DevOps work items through October 2026; phase 1 is delivered and phase 2 is in progress.
Stack
Frequently asked questions
Who is the client?
A US transportation services company with more than 150 active repositories on multi-cluster Kubernetes. We publish the case without its name.
What does Clouditive do for this transportation company?
Platform engineering, DevOps and SRE: a Backstage developer portal, self-service infrastructure with Crossplane, release engineering, GitOps with Argo CD, open-source observability, a DORA metrics programme and a shared security workflow.
How much did deployment frequency change?
Production deployments per day are roughly 3 times what they were in December 2025, measured across the whole engineering organization we support.
How did observability spend fall?
Metrics and logs moved from a commercial APM to OpenTelemetry with Mimir, Loki, Tempo, Pyroscope and Grafana, which cut spend by about USD 5,000 a month.
What is in the developer portal?
A Backstage catalog with about 66 repositories and 73 components, 11 self-service templates, golden paths for five stacks and more than 35 backend plugins.
What does build-once promotion mean?
The image is built once and the same image moves from environment to environment, so production runs exactly what was tested. It now covers more than 31 repositories, up from 4.
How large is the team?
Six people since December 2025: a lead and architect, 2 platform engineers, a DevOps engineer, an SRE and a scrum master.
What is the stack?
Multi-cluster Kubernetes, Backstage, Crossplane, Argo CD, OpenTelemetry, Mimir, Loki, Tempo, Pyroscope, Grafana, OpenCost and CodeQL.
Tell us your case.
We reply to every request within 1 business day. We sign an NDA before the call if you ask.