Infrastructure and operations
Servers, networks and services, from the network plan to the monitoring that wakes someone at night. The company runs its own fleet of cloud and on-premise machines, so I know what a system really costs over the years.
Proxmox, Kubernetes, Docker, Traefik, Google Cloud (GKE, BigQuery), AWS (Lambda, RDS, Transcribe, Polly), Azure (VMs, Azure OpenAI, AI Search), Linux, Cloudflare
Development
Web applications, interfaces and data pipelines, built so they can still be operated and changed later. Tests, CI/CD and documentation someone else can read are part of the work.
PHP / Laravel, TypeScript / Node.js, Python / FastAPI, Go, React / Next.js, PostgreSQL / MSSQL, Elasticsearch, NATS, Microservices, Events, REST / GraphQL
AI integration and LLMOps
Language models where they replace repetitive manual work: voice and telephony agents, extraction from documents and websites, content produced from templates — integrated into existing systems rather than next to them. With model serving, pipelines, monitoring and a clear cost per transaction instead of a pilot that never ends.
OpenAI / Anthropic / Gemini API, RAG / pgvector, Multi-Agent-Orchestrierung, LLMOps / Model Serving (vLLM) / Feature Stores, DialogFlow / LivePerson / NICE CXone (CCaaS), Voice / Telefonie (Asterisk ARI & AudioSocket, SIP, Twilio, Retell AI), Playwright
Migration and consolidation
Bringing a landscape that grew across several providers back together: moving services, ending contracts, unifying operations. Most recently services from several providers moved onto own hardware — while running, without downtime.
Proxmox / LXC, CI/CD (Forgejo, GitLab), Grafana / Loki / Prometheus, Backup & Recovery, Registry & Images