KAMRADEK IT-BERATUNG
From UNIX to AI
20+ years of UNIX & automation. Now: AI engineering that holds up in production.
In 1987, at fourteen, my school internship put me in front of an IBM system — a computer the size of a wardrobe, more expensive than a new car, fed with 8-inch floppy disks. One question has stayed with me ever since: how do you get machines to do the work?
For more than two decades I have been answering it as a freelance IT consultant for well-known corporations — banks, exchanges, insurers, IT service providers, and the public sector. My foundation is UNIX and Linux systems in business-critical environments, from Solaris to Red Hat, complemented by many years of network and data-center experience.
The common thread of my work is automation: for 20 years I have been teaching system landscapes to manage themselves — first with Bash, later with tools like Puppet and Ansible, and over the past ten years as a DevOps engineer in the container world with OpenShift, Podman, and CI/CD pipelines. Manual work is for things you do exactly once.
To me, large language models are the logical continuation of this path: automation that no longer just manages configuration, but understands work. Anyone who has been responsible for production systems for decades treats an LLM as exactly what it is — a powerful service that needs monitoring, access control, and a plan B.
Services
- Agentic Engineering — complete, production-ready applications, built by directing AI agents rather than hand-writing code — from Linux CLI tools to websites and mobile apps. This very website came about that way, for instance — and, as two of the many ideas I've been collecting for years, a complete algorithmic trading system and a trading-analytics platform. In the traditional world I could never have realized them all; now they finally become real — with the operations mindset that makes them dependable.
- LLM Integration — connecting large language models to your processes and systems: making internal docs, tickets, and code searchable and answerable via RAG, wiring up existing tools and APIs — from the first integration to a production-ready service.
- AI Agents — agentic workflows that take over recurring work: creating and reviewing pull requests, tests and CI/CD steps, routine tasks across development and operations — with clear guardrails instead of a black box.
- Local LLMs & Data Privacy — on-premise inference for sensitive data: model selection and hardware sizing for your environment, so that code, customer data, and operational know-how never leave your premises.
- AI-Assisted DevOps & Operations — your infrastructure work, accelerated by AI: generating and modernizing Infrastructure as Code, automating cloud and container environments, producing docs and runbooks, security reviews — log analysis and incident triage are part of it, not the whole.
- Team Enablement — getting your experienced people productive with AI tools: no hype, realistic expectations. For me this happens naturally in everyday work — over the years I've got quite a few seasoned professionals excited about it, and I'm still amazed each time by what they produce.
- Manual mode — UNIX, Linux, networking, IaC, and containers: I still know them from when you typed everything yourself. Happy to keep at it on request — but, between us: judgment should stay human; these days, the grunt work should go to the machine. Faster, and it never sleeps.
contact in the terminal.Foundation
More than two decades of project work, backed by certifications and continuous training:
- UNIX/Linux — Solaris, Red Hat, SUSE, Debian; operating business-critical environments
- Automation & Infrastructure as Code — Bash, Python, Puppet, Ansible
- DevOps & Containers — OpenShift, Podman, Jenkins, Git, CI/CD pipelines
- Networking — routing (up to BGP), load balancing, monitoring, data-center operations
- Cloud — AWS architecture and operations, plus several on-premise cloud environments
By the way: the terminal window here is real. Type help — or ask fortune for advice.
