How to Deploy a Standalone VLM Kiosk on NVIDIA Jetson with Codex

How do you deploy a standalone VLM kiosk on NVIDIA Jetson with Codex? This tutorial uses Codex to deploy a Gemma 4-powered vision-language application with Live VLM WebUI, a USB webcam, prebuilt Docker containers, and local inference on Jetson. Building on the setup from Part 1, including NVIDIA Jetson Device Skills, the walkthrough prepares the application, validates live webcam inference, and configures automatic full-screen startup. A final reboot test verifies that Jetson launches the application, reconnects the camera, and resumes local inference with the Mac disconnected. The result is a working standalone VLM kiosk that runs locally without depending on the cloud. ➡️ Start with Part 1: https://www.youtube.com/watch?v=CLGaG0JeUBI&feature=youtu.be ➡️ Learn more: https://www.jetson-ai-lab.com/tutorials/ai-assisted-development-on-jetson/ 📝 Tech blog: https://developer.nvidia.com/blog/nvidia-jetpack-7-2-1-adds-agentic-video-skills-and-t3000-emulation/ Ready to build your own? Explore models and tutorials at Jetson AI Lab.Secondary resource: https://developer.nvidia.com/embedded/learn/get-started-jetson-orin-nano-devkit 0:00 Introduction 0:14 Plan the VLM Demo 1:20 Deploy and Validate the Application 3:20 Configure a Standalone Kiosk 5:30 Reboot and Verify Automatic Startup 5:59 Standalone Demo: Local Inference, No Mac Required