Physical AI: The New Frontier

Abstract

LLMs keep improving, but scaling alone runs into real limits when you need grounded physical understanding, long-horizon planning, and reliable action, especially outside clean, text-only environments. This talk explores what’s next: world-model approaches that learn from video and interaction, multimodal perception (vision plus time-series/sensors), and agentic & multi-agent systems that connect models to tools and devices. Together, these building blocks point to Physical AI, AI that can perceive, reason, and act beyond chat in the real world.

Bio

Przemek Pospieszny is an AI Director at EPAM with 18+ years designing and implementing innovative AI solutions for global organizations. He specializes in agentic and generative AI, physical intelligence, and complex/compound AI architectures— bridging advanced R&D with production delivery. He has led and delivered dozens of AI/ML and data science initiatives for leading enterprises across Europe, the UAE, and the US. Previously, he worked in Silicon Valley with Google ATAP on human–machine interaction and sensor-driven ambient intelligence.