On-Device AI: Present State and Practical Strategy
Current State and Future of On-Device AI Technology
Model Optimization and TinyML: Practical Edge AI Strategies
AI no longer resides in the cloud alone.
Within realistic constraints of latency, security, cost, and power consumption,
AI is rapidly transitioning toward a structure where devices make direct decisions and respond independently (On-Device).
However, many organizations still hold the perception that "on-device AI is difficult" and "high-performance hardware is required."
In this webinar, we examine
optimization techniques for deploying high-performance AI models on-device, and
TinyML development methods that you can start immediately even in 32-bit MCU environments, as an integrated workflow.
💡 Key Webinar Topics
(1) AI Model Optimization Strategy with Hardware Understanding
✔️ Structural issues between explosive AI model growth and hardware performance limitations
✔️ Necessity of hardware-aware optimization beyond simple model compression
✔️ Differences between structural pruning, operator conversion, and post-training quantization techniques
✔️ Performance variance across CPU·NPU·MPU environments and optimization approaches by hardware type
✔️ Real-world cases of running generative AI·VLM·LLM on edge and on-device environments
(2) MCU-Based On-Device AI Starting with TinyML
✔️ Cloud AI vs on-device AI architecture comparison and application criteria
✔️ TinyML concepts and AI operation principles in ultra-low power and low memory environments
✔️ Lightweight model design strategies based on quantization and pruning
✔️ TinyML generation platform workflows for creating models without code
✔️ MCU AI development and deployment methods based on Keil MDK·CMSIS-NN
👀 Recommended For
✔️ Product and technology planners considering on-device AI adoption
✔️ Developers facing AI performance limitations on edge devices
✔️ Embedded engineers considering AI applicability in MCU environments
✔️ Those wanting to consolidate model compression·optimization·deployment workflows in one session
✔️ Those seeking to understand the complete picture from TinyML to high-performance on-device AI
High-performance AI optimization and TinyML,
practical solutions for on-device AI emerge at the intersection of these two technologies.
Through this webinar, explore concrete implementation strategies for on-device AI spanning models, hardware, and development environments.