A logistics ops platform that cut deploy cycles from 4 hours to 8 minutes.

Replatformed a manual ops stack onto Kubernetes with full CI/CD, observability, and zero-downtime deploys.

SoftwareDevOps

99.99%
Production uptime
8 min
Deploy time
70%
Infra cost reduction
Zero
Downtime deploys

The challenge

Where they were when we started.

Northwind moved 40,000 shipments a day across India and the GCC, but their tech stack was a tangle of manually-deployed PHP services, a single Postgres write-master held together by cron jobs, and dashboards that were six hours behind reality. Deploys were 4-hour ordeals scheduled for Sunday night. When something broke at 2am — which it did — the on-call engineer would simply restart everything.

Our approach

How we shipped it.

01
Containerized the existing stack and shipped it onto a managed Kubernetes cluster, with proper resource limits, autoscaling, and a sane secrets-management pattern from day one.
02
Built a GitHub-Actions CI/CD pipeline with preview environments per PR, automated regression tests against a production-shaped fixture, and zero-downtime blue/green deploys.
03
Stood up a full observability stack — Datadog metrics, Loki logs, Grafana dashboards — and a real PagerDuty rotation with documented runbooks for every alert.
04
Refactored the monolithic Postgres into a primary + read-replica setup with logical replication, then carved out the highest-traffic surfaces into independent services.

Outcomes

What changed for the business.

Deploys went from 4-hour Sunday-night events to 8-minute non-events that happen 30+ times a week.
Production uptime improved from a self-reported 99.5% to a measured 99.99% across a 12-month window.
Infrastructure costs dropped 70% through right-sizing and committed-use discounts on the migration.
On-call alert volume dropped 85% after the first quarter of stabilization work.