Paulina Stopa

Unix Engineer — Sky UK Ltd

Leeds, United Kingdom · February 2017 – Present · previously Shift Leader (2018–2019)

Nine years in Sky's largest Unix team, fulfilling senior-level responsibilities — delivering reliability, automation and engineering standards across a ~10,000-host estate behind national-scale services.

• Sky Unix AI Platform: designed and demoed an end-to-end AIOps proposal for the Unix estate — ServiceNow incident ingestion, LLM triage and root-cause analysis over a RAG runbook knowledge base, and AWX/Ansible remediation behind human approval gates — delivered as four interactive demo modules plus a costed requirements report.
• Develop and maintain production automation in Python, Bash and Ansible across the estate — eliminating repetitive toil, standardising change and reducing operational risk.
• Led the migration of the central LDAP service authenticating tens of thousands of users — a critical availability dependency for the entire estate.
• Mentored engineers through the Redirect platform migration — configuration standards and design review across hundreds of Sky domains, including sky.com and skysports.com.
• Former Shift Leader with Incident Coordinator training — hands-on incident management on a national-scale platform, from detection through resolution and review.
• Own documentation and engineering standards for the wider team; drive practical AI adoption across Sky's engineering community, recognised by the Head of Enterprise Product Strategy (GenAI).

Independent AI Engineering — Autonomous Operations Platform

Self-directed engineering programme · 2024 – Present

Designed, built and operate a production-grade autonomous-operations estate on my own infrastructure — the working counterpart to the platform proposed at Sky. Everything below runs today: versioned as code, monitored, and governed.

• AI platform: 63 Docker containers on a GPU-passthrough VM (~120 containers estate-wide) — LiteLLM model gateway, Ollama self-hosted inference, a Dify RAG stack with Qdrant vector search, and speech services behind Traefik ingress; all versioned Docker Compose infrastructure-as-code with CI.
• Autonomous agents: two LLM agent harnesses running 24/7 on dedicated service accounts behind human review gates — executing scheduled operational tasks and authoring incident and postmortem reports.
• ClawBoard (author; open source, MIT): agent-operations dashboard with real-time agent status over WebSocket, a full transcript audit trail, JWT authentication and a gated production/development deployment workflow — github.com/Wadera/clawboard.
• Fleet automation: 20 Ansible playbooks and 12 roles run via self-hosted Semaphore across a two-node Proxmox HA cluster — snapshot-before-change with automated restore, per-host-class patching, automatic monitoring maintenance windows and Vault-encrypted secrets.
• Observability: Prometheus and Grafana on the AI platform, with automated monitoring maintenance windows across the fleet to keep signal fidelity high during change work.
• Evaluated and self-hosted NVIDIA DGX Spark for local AI workloads (2026).

Founder & Owner — SKYDAY

Chojnów, Poland · April 2012 – February 2017

Built one of Poland's largest voice-server platforms — 10,000+ concurrent users and 100,000+ registered accounts — with a custom CRM and billing system, on self-operated dedicated-server infrastructure serving clients including the Polish National Bank and Play.

Co-founder — Digital Technologies S.C.

Poland · November 2014 – November 2016

Among Poland's first certified commercial UAV operators — aerial photogrammetry and 3D measurement models for construction and solar-installation clients.