Computer Architecture Podcast

Ep 23: Cross-Stack Design and Tooling for Large-Scale Distributed AI Systems with Dr. Tushar Krishna, Georgia Tech

In this episode, Dr. Tushar Krishna discusses his evolution from network‑on‑chip research to designing large‑scale distributed AI systems, emphasizing the importance of cross‑stack co‑design that spans accelerators, memory hierarchies, and interconnect fabrics. He explains how predictable AI data‑flow patterns enable domain‑specific optimizations, illustrated by his tools AstroSim, Chakra, and Garnet, and projects like the flexible Mary accelerator and the Maestro data‑flow language. Krishna also highlights the challenges of balancing specialization with future‑proof flexibility as AI workloads and hardware scale up to data‑center and even space‑based deployments. His perspective blends deep architectural expertise with a forward‑looking view of AI‑centric standards and industry collaboration.