About

I am Istvan. I design the systems beneath modern AI: high-performance computing, GPU infrastructure, data platforms, model-serving stacks, and the secure cloud foundations that hold them together.

My recent work sits where AI meets hard compute. I architect HPC environments for scientific simulation and engineering workloads, including CFD, fusion research, and GPU-based bioinformatics. I also design production-oriented AI systems spanning agentic workflows, managed and open-weight models, RAG, anomaly detection, document processing, and voice applications.

The recurring challenge is not getting a model to produce an impressive demo. It is making the complete system secure, observable, cost-controlled, and useful in production. I write about those constraints without hype: what works, what fails, and what the architecture looks like after the prototype.

Earlier in my career, I worked on IBM's Roadrunner supercomputer, infrastructure at Amazon, the data warehouse behind League of Legends serving more than 100 million users, and the migration of a 5 PB data warehouse to the cloud at Amadeus.

I teach Databases & Data Management at Frankfurt School of Finance & Management.