2025 Evolution Episode 10: Introduction to AI Infrastructure: Networking and Hardware

This three-part episode examines the bandwidth, latency, nonblocking-fabric, and congestion-control demands of AI/ML workloads; introduces specialized networking technologies such as RoCE and RDMA; and maps the roles of CPUs, GPUs, DPUs, SmartNICs, NVLink, and BlueField across an end-to-end AI infrastructure.

Total video duration: 2 hr 2 min 21 sec.

Companion Git repository: No dedicated repository for this episode.