We're not tool-religious. We're outcome-religious. This is the current, honest inventory of what we run in production for clients, plus the alternatives we'll happily reach for when your context demands it.
Grouped the way we think about them internally, not the way vendors sell them.
Deep production experience across the three hyperscalers and specialist clouds. We design cloud-agnostic where it saves money, and lean into managed services when it saves time.
Kubernetes for stateful workloads, Nomad for edge and mixed fleets, Agones for real-time. Fleet upgrades treated as a first-class product.
Everything reproducible, everything reviewed. Argo CD for continuous delivery, Renovate for supply-chain hygiene, cosign for image provenance.
No clicking. Terraform + Terragrunt for the boring parts, Pulumi for typed control planes, Crossplane where the API surface belongs in the cluster.
You cannot operate what you cannot see. We standardise on OpenTelemetry and pick the backend that fits your budget and geography.
Zero-trust identity, secrets in vaults not env files, runtime detection, and audits your compliance officer will actually enjoy reading.
From single-node Postgres to petabyte lakehouses. Streaming pipelines that survive a broker outage and analytics that survive a Monday morning.
GPU platforms with real cost accounting, model serving that hits your SLA, and evaluation pipelines so you know when quality drifts.
Service mesh where it earns its keep, private connectivity across clouds, and edge presence tuned per market.
SSO consolidated, machine identity federated, blast radius controlled. Break-glass paths that are actually auditable.
Institutional-grade validator and RPC operations. HSM/MPC signing, MEV-aware relayers, archive indexing.
Right compute for the workload. Baremetal when latency demands it, spot when budget demands it, serverless when velocity demands it.
Managed services buy you speed. We swap them out only when the math clearly beats the operational cost.
OpenTelemetry, OCI, S3-compatible, Postgres wire. The exit is designed the day we enter.
Infra, apps, dashboards, runbooks, all in git, all reviewed, all reproducible in a clean AWS account.
The interesting problems live at the edges. Bring us the weird one.