Work
Six years of work: from Kubernetes clusters and build pipelines to AI platforms and corporate services.
What I have built
I built and run clusters at five companies: two in the cloud, three on their own hardware. The largest has 21 worker nodes and 3 control-plane nodes, over 700 microservices. They come up from scratch through Terraform (vSphere/ESXi) and Ansible — minutes instead of hours. The closed network is reachable only over VPN. Inside: networking and ingress (Calico, MetalLB, Ingress NGINX, cert-manager) and storage under load (NFS, Longhorn, vSphere CSI, MinIO).
A single process for building, shipping and storing releases across every team. The service’s language does not matter: werf builds the artefact from a Dockerfile, so React and Vue, Go, Python and C# live on the same pipeline — and any other language would too. Linters and code analysers, caching and automatic tagging are built into the pipelines. Releases follow GitOps practices across a multi-product estate.
A Telegram monitoring platform: it collects messages from chosen channels and chats and prepares personal daily digests with links back to the originals. Separately — more than ten AI agents for specific business tasks, wired into the company’s existing services.
A custom implementation in Go on top of CoreDNS. It spreads requests across servers with no downtime and no bottlenecks.
One way in to every internal service. Passwords and keys moved out of the code into Vault, integrated with GitLab CI; secrets migrated out of Kubernetes. Access to company assets is managed in one place.
Time-series databases in the branches, with metrics gathered in a star topology into a single monitoring and response centre inside a closed network. Open-source metric collectors were extended with client self-signed certificate authentication.
A single way in to the company’s internal services, with access under control. Alongside it, a full set of tools for remote and office work: document handling on a private cloud, a corporate chat and video calls. All of it on the company’s own servers.
Automatic health checking of public proxies. It supports seven protocols — from SSH to Xray.
Got a similar problem? Tell me about it and I will show how it could work for you.