How do you run trillion-parameter AI models and rapidly switch between multiple AI agents without hitting the traditional memory wall?

Sumti Jairath, Chief Architect at SambaNova Systems, explains how memory-centric design and RDU dataflow architecture change the way large-scale GenAI and agentic workloads are executed.

Why does agentic AI place different demands on the infrastructure running the model?

Pushkar Nandkar explores SambaNova's SN40L and explains how dataflow architecture can deliver the inference performance needed when AI moves beyond individual prompts into complex, multi-stage agentic workflows.