InfoQ talk argues reliable AI platforms need tools and agents

InfoQ published a presentation on designing AI platforms for reliability, featuring Aaron Erickson’s explanation of how NVIDIA designs and tests purpose-built AI agent hierarchies.

The talk argues that production AI systems need deterministic tools for certainty alongside agents for discovery, with LLM-as-a-judge test pyramids and careful use of rare context.

For senior developers and architects, the theme is that reliability comes from system design, not from assuming an agent will reason correctly every time.