About Me

I’m Adel Iazzag, a Principal Network Engineer focused on AI infrastructure, GPU cluster networking, InfiniBand, RoCE, Ethernet fabrics, and large-scale data center design.

My work sits at the intersection of networking, distributed systems, and AI infrastructure. I spend a lot of time thinking about how GPUs communicate, how training workloads stress the network, and how protocols like InfiniBand, RoCEv2, BGP, EVPN/VXLAN, ECN, PFC, and congestion control behave under real production pressure.

But I’m not only curious about networking.

I’m also trying to understand everything adjacent to it: GPU architecture, distributed training, parallelism strategies, NCCL, collectives, storage, compute, physical infrastructure, and the way all these layers come together to make large-scale AI systems work.

This blog is first a resource for me.

It’s where I organize what I’m learning, write things down clearly, and force myself to understand the systems I work with at a deeper level.

The second goal is to help other engineers learn the topics I struggled with. A lot of AI infrastructure and networking content is either too shallow or written like an academic whitepaper. I want this blog to sit in the middle: practical, visual, simple when it can be, and deep when it needs to be.

I focus on topics like:

• AI cluster network architecture
• InfiniBand and RoCE design
• NCCL, collectives, and GPU communication patterns
• GPU architecture and how it impacts infrastructure
• Distributed training and parallelism strategies
• Data center fabrics and Clos topologies
• BGP, EVPN/VXLAN, and large-scale routing design
• Congestion control, lossless Ethernet, and RDMA
• Physical layer design: optics, fiber, racks, power, and cooling
• Career growth for infrastructure and network engineers

My goal is to make complex infrastructure topics easier to understand without watering them down. I try to explain systems from first principles, using diagrams, packet walks, tradeoffs, and practical examples.

I’m especially interested in the infrastructure behind large-scale AI: not just how packets move, but why design choices matter when thousands of GPUs need to behave like one giant computer.

This blog is part learning journal, part technical notebook, and part attempt to help other engineers build stronger intuition around AI networking, GPU infrastructure, and data center systems.

You can also find me on LinkedIn, where I share shorter thoughts on AI infrastructure, networking, and engineering growth.