At Patrik Duch, I stabilized a high-growth US SaaS platform by optimizing AWS and Kubernetes infrastructure, resolving bottlenecks, and improving reliability to 99.999%+ uptime. I also designed a cross-region disaster recovery strategy that reduced recovery time from hours to minutes.
For the Czech Ministry of Labour’s ePodatelna system, I architected a cloud-native training sandbox supporting 10.9M+ citizens and simulating 50,000+ onboarding events per day. I also prototyped an LLM retrieval pipeline using adaptive chunking, LangGraph, and LangFuse.

