From Footfall to Behavior: Tracking Behavior Without Tracking People inside Shopping Malls
Mon Jan 12 2026
In this episode of the podcast, I sit down with Galvin Widjaja, CEO of Lauretta, to unpack what it really means for computer vision AI to “understand” the physical world. Moving beyond hype around models and automation, the conversation explores why most computer vision systems fail in real environments touching on dataset limitations, time and identity problems, and the difference between counting people and understanding behavior. Drawing on real-world deployments with organizations such as Lendlease (shopping mall), Changi Airport, the Transportation Security Administration (TSA), and the U.S. Department of Homeland Security (DHS), Galvin explains why effective physical AI isn’t about giving machines human-like intelligence, but about designing systems that detect patterns, respect privacy, show anomalies, and leave judgment and action to humans who have skin in the game.
More
In this episode of the podcast, I sit down with Galvin Widjaja, CEO of Lauretta, to unpack what it really means for computer vision AI to “understand” the physical world. Moving beyond hype around models and automation, the conversation explores why most computer vision systems fail in real environments touching on dataset limitations, time and identity problems, and the difference between counting people and understanding behavior. Drawing on real-world deployments with organizations such as Lendlease (shopping mall), Changi Airport, the Transportation Security Administration (TSA), and the U.S. Department of Homeland Security (DHS), Galvin explains why effective physical AI isn’t about giving machines human-like intelligence, but about designing systems that detect patterns, respect privacy, show anomalies, and leave judgment and action to humans who have skin in the game.