Currently booking Q3 2026 engagements
We're a small team that builds and operates Kubernetes, AI, and automation for companies in the messy middle. Production-grade, cost-aware, and handed off so you're not stuck with us.
NAMESPACE NAME READY STATUS AGE
prod api-gateway-7d8f9b-2k4lm 1/1 Running 12d
prod inference-vllm-0 1/1 Running 3d
prod rag-retriever-5c7b8d-x9pq2 1/1 Running 18h
prod embedder-bge-0 1/1 Running 18h
data vector-db-shard-0 1/1 Running 31d
infra prometheus-server-0 2/2 Running 62d
infra grafana-78b6f-mz4cv 1/1 Running 62d
infra cert-manager-849-h7rqx 1/1 Running 85dStacks we ship on
Services
We don't do everything. We do infrastructure for tech companies that have found product-market fit and are starting to feel the weight of it. The three practices below are how that usually shows up.
01 / Kubernetes
We design, deploy, and operate Kubernetes for teams that need to scale without inheriting the operational debt. EKS, GKE, self-hosted — we've moved companies from Compose to multi-region production in weeks, not quarters.
Cluster design · Migrations · CI/CD · Monitoring · Cost optimization
02 / AI
Moving AI from a Jupyter notebook to a production system that survives Monday morning traffic is a different skill set. We build the inference, retrieval, and cost-control layers around the model — whether it's OpenAI, Anthropic, or your own fine-tunes.
LLM APIs · RAG · Model serving · Inference optimization · Cost controls
03 / Automation
Terraform, Ansible, and GitOps that your team will actually maintain after we hand it off. No bespoke abstractions or DSLs no one can read — just clean infra-as-code your engineers can navigate on day one.
Terraform · Ansible · GitOps · Documentation · DR & backups
Selected work
Names are kept off the page; we'll share details under NDA when it's relevant to your problem.
Series B fintech
Migrated a payments API serving 4M req/day onto multi-region EKS with cross-region replicated Postgres. Cutover with zero customer-visible downtime.
AI startup
Built the retrieval, evaluation, and cost-control stack for a vertical AI product. Self-hosted embedding model, OpenAI for generation, full token-budget guardrails.
Enterprise media co.
Replaced a Jenkins maze with ArgoCD + Terraform Cloud. Trained the platform team and stayed for two months of hand-off support, then we left.
How we work
Engagements are scoped, time-boxed, and structured so your team owns the result when we're done.
01
We start with the real questions: what's breaking, what's slowing you down, what's keeping your engineers up at night. Not a sales call — an investigation.
02
No bespoke abstractions or hand-rolled DSLs. You get Terraform, Kubernetes manifests, runbooks, and architecture diagrams your engineers can navigate on day one.
03
We deploy, set up monitoring, train your team, and stay for the edge cases. Then we actually leave. You're not stuck with us forever.
We typically work with tech companies that have figured out product-market fit and are starting to scale operations. Not the MVP stage. Not the already-tidy stage. The messy middle.
Team size
10 – 200 engineers
Stage
Post-PMF, pre-scale
Engagement length
2 weeks → 3 months
Get in touch
Whether it's Kubernetes chaos, AI deployment, or just expert advice on your architecture — we'll ask good questions and give you honest answers.