Infrastructure that stays boring
Cloud & Platform Engineering
Reproducible, observable, cost-controlled infrastructure — on the cloud, in your data centre, or both, with the migration path between them.
- Cloud spend reduction
- 38%
- Median saving in the first six months after a cost-engineering review.
- Mean time to recovery
- 82%
- Improvement after introducing tracing and automated rollback.
- Environment provisioning
- 45 min
- Time to stand up a complete, isolated environment from an empty account.
Our position
How we think about it
Good infrastructure is unremarkable: it deploys the same way every time, it tells you when something is wrong before a customer does, and the bill does not surprise you. We build exactly that. Everything is defined as code, every environment is reproducible from an empty account, and every service emits the traces and metrics needed to answer "why is it slow" without guessing.
What you receive
- Target architecture with cost model and failure analysis
- Fully version-controlled infrastructure
- Environments reproducible from scratch in under an hour
- CI/CD with automated rollback on error-budget burn
- Observability stack with meaningful, low-noise alerts
- Documented and rehearsed disaster-recovery procedure
- Monthly cost report with rightsizing recommendations
Capabilities
What we actually build
- Cloud architecture on AWS, Azure, Google Cloud and Hetzner
- Infrastructure as code with Terraform and Pulumi
- Kubernetes platform engineering and internal developer platforms
- CI/CD pipelines with progressive delivery
- Observability — metrics, logs, traces and alerting that people trust
- Cost engineering and rightsizing
- Disaster recovery design and rehearsal
- Data-sovereignty compliant hosting inside South Africa
Technologies we use
- Terraform
- Kubernetes
- Docker
- ArgoCD
- GitHub Actions
- Prometheus
- Grafana
- OpenTelemetry
- Cloudflare
- AWS
- Azure
- Hetzner
Process
How a cloud & platform engineering engagement runs
Each phase ends with something you can look at and a decision you can make.
- 012 weeks
Assess
Current architecture, spend, failure modes, recovery objectives and compliance constraints — measured, not assumed.
- 022 weeks
Design
Target state with an explicit cost model and a failure analysis for every tier.
- 033–6 weeks
Codify
Everything becomes Terraform. If it was clicked in a console, it does not exist.
- 044–16 weeks
Migrate
Service by service, with traffic shifting gradually and a tested route back at every stage.
- 05Ongoing
Operate
Error budgets, on-call rotation design, quarterly DR rehearsal and continuous cost review.
Questions
Cloud & Platform Engineering — common questions
If yours is not here, ask Nova or send it to us directly.
Does our data have to leave South Africa?
No. AWS Cape Town, Azure South Africa North and several local providers give you full in-country residency. We design for POPIA data-residency requirements as a starting constraint, not a retrofit.Do we need Kubernetes?
Probably not, and we will say so. Plenty of workloads run better and cheaper on managed containers or a couple of well-configured virtual machines. We recommend Kubernetes when the operational complexity actually buys you something.Can you work with our existing DevOps team?
Yes. Most cloud engagements are joint — we bring the platform patterns and the automation, your team brings the domain knowledge, and the goal is that they own it when we leave.What about hybrid and on-premise?
Common in mining and government, and fully supported. The same Terraform and Kubernetes patterns run on your own hardware, with cloud used selectively for burst capacity or disaster recovery.