Cloud Craft helps operators and product teams design, automate, monitor, and maintain infrastructure that needs to stay online — from Linux and Kubernetes platforms to AWS, Azure, Google Cloud, private cloud, hybrid environments, networking, storage, and custom operator software. We work across DevOps, SRE, infrastructure architecture, and broadcast / IPTV systems. NoSignal, a product we build and operate, is one example of how we work.
About
Cloud Craft is run and staffed by senior engineers. We work on systems where the cost of downtime, weak automation, bad observability, fragile networking, or poor cloud architecture turns into real operational pain — paged engineers, missed SLAs, slipped migrations, and stakeholders asking hard questions. Our recommendations assume someone will be acting on them at 3 a.m. with limited context, and we design for that case.
We design across Linux infrastructure, public and private cloud, Kubernetes, networking, and observability. We deploy to real production environments — bare metal, virtualized, AWS, Azure, GCP, on-prem — and we instrument with Prometheus and Grafana from the first day. We treat configuration as code that has to survive an on-call rotation. We prefer small, deterministic systems that can be reasoned about under stress over large, clever ones that cannot. None of this is unusual — it is the work, done explicitly.
Services
Six practice areas. Engagements often span more than one — infrastructure, cloud, networking, observability, and software usually fail or succeed together.
Linux, systemd, nginx and HAProxy, CI/CD pipelines, build and release automation, secrets management, runtime hardening, and infrastructure automation. The unglamorous half of production — the part that decides whether your platform is reliable or merely lucky.
Cloud architecture, migrations, private cloud platforms, hybrid connectivity, and cost-aware deployments. We design cloud environments that can be operated clearly, secured properly, and moved forward without locking the whole business into fragile assumptions.
Cluster design and operation, ingress, autoscaling, RBAC, GitOps, deployment workflows, and internal developer platforms. Built so application teams can ship features without taking on the infrastructure underneath.
Routing, BGP and MPLS knowledge, load balancing, HA design, storage, multipath, and virtualization with Proxmox and KVM. For environments where “best effort” is not acceptable and the network is part of the product.
Prometheus, Grafana, log pipelines, metric design, alerting, SLOs, dashboards, and runbooks. Without observability, every incident is a guess and every fix is a hope.
Rust, Go, Python, and TypeScript. REST and SSE APIs, internal tools, control planes, automation tools, and streaming and media systems (including MPEG-TS where relevant) — written by people who run them in production.
Projects
Cloud Craft develops and operates its own products. They are not portfolio pieces — they are real software built for operational environments. NoSignal is the clearest example of how we approach infrastructure-adjacent software: Linux-native, observable, deterministic, explicit, and operator-focused.
NoSignal is deployed in production by broadcast and IPTV operators in the EU. It reflects Cloud Craft's engineering style: Linux-native, observable, deterministic, configured explicitly, and built around what operators actually need on a busy day.
Why us
A short list of what to expect when you work with us. None of it is a slogan. Each is a constraint we work under. We combine skills that are usually split across several vendors — cloud architecture, Linux operations, DevOps and SRE, networking, virtualization and storage, observability, software development, and broadcast / media infrastructure.
The people who design your infrastructure and write your code are the same people you talk to. There is no junior layer translating decisions into output.
Every engagement is built from your platform's actual constraints. Nothing is copy-pasted from a previous client. Boilerplate exists only where boilerplate belongs.
Decisions are made for what runs at 3 a.m., not for what demos well. We design for the steady state and the worst day, not for the first impression.
Metrics, logs, alerts, runbooks, and deployment automation are part of the work — not items added later, after something has already broken.
A surprising amount of the job is figuring out what is actually happening. We do that work explicitly, with measurements, not guesses.
Public cloud, private cloud, Linux infrastructure, networking, virtualization, storage, observability, software development, and broadcast / media systems — informing each other, not split across separate vendors.
Contact
Send us the system, migration, reliability, cloud, networking, or automation problem you are dealing with. We will tell you where we can help — and where we cannot.