Keynote: Leveraging Zephyr and ML to Bring Smart Devices to Market, Faster
summary
Zephyr integrates on-device ML using TensorFlow Lite Micro plus microTVM and Edge Impulse, with graphics via LVGL, and automated SBOMs in SPDX formats. The project shows scale: 100,000 commits and about 700 boards supported, 220 sensors, and 2.8 commits/hour. Concrete products include a Moto watch, Samsung Galaxy Ring, an iris scanner, grid monitors like Ram One designed to last five years on one battery, trash-can image detectors, implantable anti-poaching sensors, and a tiny‑ML “nose” successfully ported to Zephyr.
AI-written summary · about
chapters (11)
- 0:00Zephyr and smart devices
- 2:06Zephyr’s device role
- 4:14Zephyr’s popularity
- 5:42Smart device requirements
- 7:09Hardware and sensors
- 8:13Zephyr AI/ML integration
- 9:18Graphics and user interface options
- 10:02Memory and power optimization
- 10:44Zephyr security
- 12:07Real-world case studies
- 13:52Resources and community
related talks
10mLightning Talk: Debugging Edge AI on Zephyr and Lessons Learned
36mTinyML at the Edge: Deploying and Optimizing AI Workloads on Zephyr RTOS
01h 16mOpening Keynote
07mLightning Talk: Zephyr Portability with an AI Application on Very Different MCUs
39mFrom a Weekend Hack, to a Future-Proof IoT Device
01h 19mMini-conference: RISC-V Support in Zephyr
42mPorting an AI Powered Wearable Health Monitor to Zephyr on Open Hardware
31mKeynote: The next big thing in machine learning is tiny