Uploaded January 2026 | Updated November 2026, 2 hours ago
This video demonstrates how advanced AI/ML vision workloads can be deployed directly on resource-constrained edge devices without relying on the cloud. Using the Silicon Labs EFR32 MG26 SoC and ModelCat’s (formerly Eta Compute) Aptos platform, we showcase real-time face and gesture recognition running on an MCU with tight memory, power, and performance budgets.
The demo highlights how a dedicated machine learning hardware accelerator, combined with an efficient radio and low-power architecture, enables responsive vision inference at minimal energy cost. With concrete performance and power metrics, this solution illustrates how developers can bring intelligent, always-on interactions to everyday connected products like smart locks, switches, and lighting.
🔗 Learn more: silabs.com/applications/artificial-intelligence-machine-learning?source=Social&detail=YouTube&cid=soc-you-ml-011225
This video demonstrates how advanced AI/ML vision workloads can be deployed directly on resource-constrained edge devices without relying on the cloud. Using the Silicon Labs EFR32 MG26 SoC and ModelCat’s (formerly Eta Compute) Aptos platform, we showcase real-time face and gesture recognition running on an MCU with tight memory, power, and performance budgets.
The demo highlights how a dedicated machine learning hardware accelerator, combined with an efficient radio and low-power architecture, enables responsive vision inference at minimal energy cost. With concrete performance and power metrics, this solution illustrates how developers can bring intelligent, always-on interactions to everyday connected products like smart locks, switches, and lighting.
🔗 Learn more: silabs.com/applications/artificial-intelligence-machine-learning?source=Social&detail=YouTube&cid=soc-you-ml-011225








![Hands-On Edge AI: Developing Embedded AI/ML Applications - Silicon Labs
AI/ML at the edge is rapidly transforming how IoT devices sense, interpret, and respond to the world around them. But moving from concept to a working embedded AI application can feel complex and fragmented.
This Tech Talk is designed to simplify that journey.
In this session, we will build an AI/ML application step-by-step, walking through the full development flow from model integration to deployment on an embedded device. The focus will be on practical implementation — understanding how models move from development to embedded inference, how performance and memory tradeoffs are evaluated, and how tooling supports efficient iteration.
By the end of the session, you’ll have a clearer roadmap for bringing AI/ML capabilities into your own embedded designs.
Chapters:
[0:00] Introduction
[1:53] What is ML at the edge?
[6:05] How does Silicon Labs enables the IoT?
[6:45] ML at the edge application examples
[11:45] ML software pipeline
[12:55] Partner software and tool support
[17:00] Glass break reference design for security applications
[18:21] Gesture detection with ModelCat AI
[20:01] BLE positioning with ML with Unikie
[23:50] Machine learning voice controlled light demo
[34:04] Silicon Labs ML portfolio and kits
[39:48] Resources
Further information:
- Tech Talks slides: https://www.silabs.com/support/training/hands-on-edge-ai-developing-embedded-ai-ml-applications?source=Social&detail=YouTube&cid=soc-you-all-111126
- IoT Machine Learning Solutions: https://www.silabs.com/applications/artificial-intelligence-machine-learning?source=Social&detail=YouTube&cid=soc-you-ml-111126
- AI/ML Developer Journey: https://www.silabs.com/applications/artificial-intelligence-machine-learning/ai-ml-developer-journey?source=Social&detail=YouTube&cid=soc-you-ml-111126
#MachineLearning #AIML #edgeAI #IoT #TechTalks Hands-On Edge AI: Developing Embedded AI/ML Applications - Silicon Labs](https://i.ytimg.com/vi/yE8y96tRJuU/mqdefault.jpg)

