About

I’m Tom Pechulis, an HPC and AI infrastructure architect. I design and build large GPU clusters for AI training and inference, from NVIDIA HGX systems through GB200 and GB300 NVL72 racks.

This blog is where I write about how that infrastructure actually comes together: the compute, the fabric, the power and cooling, and the economics that decide what gets built. The goal is practical and specific, the kind of writing I wanted when I was learning this.

I studied Computer Science, with a second major in Aviation Science, at Baylor. Outside of work it’s aviation, home-lab projects, and real estate.

Views here are my own and don’t represent any employer, customer, or vendor.

More about me at tompechulis.com. New posts are on the RSS feed.