Sensor-Driven Physical AI: Modern Solution Design Patterns for Edge Intelligence

Session details:

In this talk, we’ll share a set of solution design patterns that have repeatedly proven effective across real edge AI deployments. These patterns aren’t theoretical—they’ve emerged as teams optimize for the realities of edge intelligence: tight power and memory budgets, noisy sensors, intermittent connectivity, safety and privacy expectations, and the need to scale from a prototype to a fleet.

You’ll see how modern systems combine lightweight sensing and perception with selective escalation to richer multimodal reasoning (including vision‑language models), and how agentic workflows can turn sensor signals into coordinated actions—with guardrails—rather than dashboards and alerts.

Walk away with a clearer mental model for designing edge systems that don’t just produce predictions, but deliver outcomes—fast, private, and dependable—where the world actually happens.