— Practice / 02 / Foundations

Cloud Engineering.

AI workloads punish poor foundations. We design and build the cloud substrate that scales with the leap — secure by construction, observable by default, economical at the scale your models demand.

§ 01
Why this practice exists

The cloud bills of the AI era are not the cloud bills of the SaaS era. Inference rewrites the unit economics of every workload it touches.

Most cloud estates were sized for predictable, mostly-stateless services. AI estates need GPU density, low-latency retrieval, ruthless cost discipline, and observability that goes deeper than logs.

Our cloud practice is not commodity. It is the foundation that decides whether your leap actually lands.

§ 02
Capabilities

What we build.

From greenfield landing zones to deeply-entangled modernisation programmes — built to a single bar.

— 01

Infrastructure Modernisation

Lift-and-shift considered harmful. We re-architect for elasticity, governance, and the compute density that inference workloads require — across the long-tail estate, not just the green-field.

ContainerisationIaCCompute
— 02

Landing Zones & Multi-Cloud

Account hierarchy, network topology, identity, guardrails. AWS, Azure, GCP — designed for the workloads, not the brochure. Built to pass the audits you have not had yet.

AWSAzureGCP
— 03

Platform Modernisation

Kubernetes, service mesh, secrets, GPU-aware scheduling, ingress, certificate lifecycle. The substrate that makes everything else possible — and stays out of the team's way.

KubernetesMeshGPU scheduling
— 04

Cost & FinOps Discipline

Token economics, GPU rationalisation, commitment shape, autoscaling fences. AI bills are the new headline risk; we design them down — without mortgaging future flexibility.

FinOpsToken costGPU/TPU
— 05

Security by Construction

Zero-trust patterns, workload identity, supply-chain attestation, secrets, KMS hierarchy. Compliance becomes a side effect, not a project.

Zero trustSBOMKMS
— 06

Observability & Reliability

OpenTelemetry-first telemetry, SLOs that mean something, error budgets that get respected. The plumbing that makes 24×7 operations possible.

OTelSLOTracing
§ 03
Where we tend to live

A pragmatic stack.

We are technology-agnostic, not technology-naïve. This is the broad surface our partners spend most of their time on.

— Compute & orchestration
Kubernetes (EKS, AKS, GKE), Karpenter, KEDA, Nomad, Bottlerocket
— Networking & mesh
Istio, Linkerd, Cilium, AWS Transit Gateway, Cloudflare
— Identity & secrets
IAM, Workload Identity, Vault, External Secrets, OPA, Cedar
— IaC & pipelines
Terraform, OpenTofu, Pulumi, Crossplane, Argo, Flux
— Observability
OpenTelemetry, Grafana, Prometheus, Datadog, Honeycomb
— FinOps & cost
CUR, Cost Explorer, Vantage, Kubecost, custom unit economics
The cloud is no longer the leap.
It is the floor.
— Cloud Engineering at Revoleap
§ 04
Adjacent practices
— 01 / Practice in chief
AI Engineering
Read →
— 03 / Substrate
Data
Read →
— 04 / Surface
App Engineering
Read →
— 05 / Always-on
Platform Ops
Read →

Architect the substrate.

A 90-minute working session with our cloud principals — landing-zone review, cost benchmark, GPU readiness.

Begin a conversation
§ 04
Sample artifact · how a cloud TOM lays out
— Figure · a target operating model, abstracted

The shape a cloud function takes when it's working.

— STRATEGY Cloud as advantage — ORG Platform team SREs, architects, FinOps, security — PROCESS Engineering loops IaC · CI/CD · observability incident response · reviews — PARTNERS Hyperscalers GCP · AWS · Azure + delivery firm — COMPUTE Kubernetes, serverless, GPU — DATA Lakehouse, streams, vector — NETWORK VPC, edge, private peering — SECURITY IAM, secrets, audit, posture — PLATFORM

— The strategy at the top is decided by the leadership team. The operating model in the middle is what most transformations under-engineer. The platform at the bottom is where engineering teams spend 80% of their time. We work from the top down and the bottom up at the same time.