Aurora LMS
Kindle Academy's LMS was falling over at exam time. We rebuilt it as one platform for 80,000 students.
INFRASTRUCTURE AS A PRODUCT
Cloud & DevOps Engineering, Scalability & SRE
We run infrastructure like an engineering product: infrastructure as code, GitOps delivery on Kubernetes, automated CI/CD and SRE practice. When the product is already slow or old, we find the bottleneck and migrate without downtime.
Terraform IaC, Kubernetes platforms and GitOps delivery as standard
We start from the slow query and the 9am traffic spike, not a diagram
Zero-downtime migrations: add, dual-write, backfill, then switch
HOW WE BUILD THIS
Infrastructure you can change without a hero weekend
Reliable cloud operations are boring by design: every change reviewed and automated, every rollback rehearsed, every slowdown measured before anything is touched. We build that operating model, then fix what is actually hurting.
What this looks like in the codebase
After the workflow is clear, we lock the technical choices that keep the product maintainable. These are the ones we use here:
- Kubernetes (EKS/GKE) platforms with GitOps delivery via ArgoCD or Flux
- CI/CD pipelines with preview environments, automated tests and one-click rollbacks
- SRE operating model: SLOs, error budgets and golden-signal dashboards
- PostgreSQL performance: query plan analysis, partial indexing and read replica pooling
- Zero-downtime migrations: dual-writes, backfills and staged switchovers
Stack we ship with
Chosen because we have run it in production, not because it is fashionable.
AWSGoogle Cloud PlatformKubernetes (EKS / GKE)TerraformArgoCD / FluxGitHub ActionsPostgreSQLRedisKafka / RabbitMQDatadog / OpenTelemetry
WHAT YOU GET
The work inside this service
From the first data model to a production deploy. Here is what we hand over.
01 / CLOUD & DEVOPS ENGINEERING
Kubernetes & Infrastructure as Code
Clusters, networks and cloud accounts defined in Terraform and delivered with GitOps. No console-clicking, no drift, no snowflake environments.
- EKS/GKE platform provisioning with Terraform
- GitOps delivery with ArgoCD or Flux
- Secrets management and policy guardrails (OPA)
- Cloud cost controls, rightsizing and tagging standards
02 / CLOUD & DEVOPS ENGINEERING
CI/CD Automation & SRE Operations
A pipeline from commit to production that tests, previews and rolls back by itself — with the dashboards and on-call practice to run it calmly.
- CI/CD pipelines with preview environments
- Automated tests, migrations and rollbacks in the pipeline
- SLOs, error budgets and golden-signal dashboards
- Alerting, on-call runbooks and incident reviews
03 / CLOUD & DEVOPS ENGINEERING
Scalability, Performance & Zero-Downtime Migration
Slow queries, exhausted pools and old systems. We read the query plans, fix the data path, and move the old system in steps while the business stays up.
- EXPLAIN ANALYZE optimization and index design
- Caching layers and connection pooling (PgBouncer, Redis)
- Strangler-fig and dual-write migration roadmaps
- Disaster recovery and automated backup validation
WHY THIS HOLDS UP
A call we made in production
Reducing P99 Query Latency from 1.8s to 32ms Under 40,000 RPM
A CALL WE MADE IN PRODUCTION
Reducing P99 Query Latency from 1.8s to 32ms Under 40,000 RPM
When auditing a high-volume logistics dispatch dashboard suffering from database connection starvation during peak operating hours, we diagnosed unindexed sequential table scans joining 6 relational tables on every polling cycle. We restructured the read pipeline: introduced Redis read-through caching for volatile active driver locations, implemented partial composite indexes on active dispatch statuses, and deployed PgBouncer connection pooling. P99 query latency dropped from 1,840ms to 32ms while reducing database CPU utilization from 94% to 18%.
WHAT CHANGED
View all case studies Products we built this way
Ledgerly
Strata Retail spent 35 hours a week reconciling supplier invoices by spreadsheet. We built one ops portal for 24 branches.
Northbeam Ops
Dispatchers were refreshing a 50-column spreadsheet while 350 trucks pinged GPS. We built a live dispatch board that keeps up.
FREQUENTLY ASKED QUESTIONS
Questions before you write to us
Yes. We audit the Terraform, the clusters, the pipeline and the alerting first, then take over with a written operating model: what is monitored, what happens on page, and what the rollback looks like.
Yes. We start with query plans, connection pool use and cache hit rates. Then we change indexes, add a cache, or split a hot path. A rewrite is last, not first.
The plan is: add compatible columns, write to both, backfill in chunks, switch reads, then drop the old columns. We do not flip a switch on Friday night and hope.
We leave you with dashboards for latency, errors and database health, plus alerts when a pool or a query goes bad. We stay available for the next change.
Ready to turn this into a product? Tell us what you need.