01 AI infrastructure
02 Which one
| You are | Start with | Because |
|---|---|---|
| Spending more on model APIs each month and unable to explain why | AI Infrastructure Setup | Week one is measurement. Even if you never self-host anything, you end up knowing what you spend and on what — and the gateway makes that permanent. |
| Already self-hosting, and it has quietly become nobody's job | AI Ops Retainer | We take over existing stacks. First month is an audit you keep regardless of what you decide afterwards. |
| Being asked by legal, security or a customer where the prompts go | AI Governance | The controls need infrastructure to live in. We build the evidence; your counsel makes the argument. |
| Operating in or entering the China market | China & APAC | Connectivity, ICP filing and acceleration are prerequisites, not add-ons — including for inference that has to answer inside the mainland. |
| Without a cloud footprint worth putting GPUs on yet | Cloud foundations | Accounts, identity, network and budgets first. Everything else inherits from that layer or routes around it, and routing around it is how estates rot. |
03 China & APAC
The practice that built this company.
Cross-border connectivity, ICP filing, acceleration and mainland access work, running since 2017. Still sold, still supported, still the reason a decent number of our clients found us.
04 Cloud platforms
Six platforms, no loyalty to any of them.
Migrations, landing zones, Kubernetes, Terraform and the ongoing operation of all of it — which is now mostly in service of putting inference somewhere sensible.
Cloud migrations
Between providers, between regions, out of a datacentre, or off a vendor API onto your own hardware — increasingly the last one.
Kubernetes
Cluster design, GPU node pools, autoscaling, and the operational discipline that keeps a cluster from becoming the problem it was meant to solve.
Platform engineering
The long-running arrangement: we are the platform team you have not hired, for the parts where that makes economic sense.