Uploaded August 2026 | Updated September 2026, 1 week ago
SPEAKERS:
Niraj Dengale, Andes Technology Co. Ltd.
Frank, Yueh-Feng Lee, Andes Technology Co. Ltd.
This session was a part of the Software track at the RISC-V Developer Workshops held at RISC-V Summit Europe 2026 in Bologna, Italy.
WORKSHOP DESCRIPTION:
Hands-on TinyML Deployment on RISC-V: Building an Optimized Edge AI Pipeline - This hands-on workshop introduces a practical workflow for deploying TinyML applications on RISC-V edge platforms, demonstrating how developers can move from neural network models to efficient hardware-accelerated inference.
Participants will explore a complete TinyML deployment pipeline using an open RISC-V AI software stack. The session focuses on how neural network models designed for resource-constrained devices can be optimized and executed efficiently on RISC-V processors using optimized kernel libraries.
In collaboration with Edge Impulse (pending confirmation), attendees may explore how TinyML models can be developed and exported for embedded deployment using the Edge Impulse platform. If this collaboration is not available, the workshop will instead demonstrate a similar workflow using model preparation and optimization tools provided by Andes Technology.
The optimized models will then be deployed using neural network kernel libraries from Andes Technology, demonstrating how operator-level optimizations and efficient memory access patterns improve inference performance on RISC-V processors.
Example workloads may include representative TinyML models from MLPerf Tiny. These models illustrate how model architecture, kernel implementations, and RISC-V compute extensions such as vector or DSP instructions interact to deliver efficient edge AI inference.
Through this guided hands-on session, developers will gain practical experience deploying AI workloads on RISC-V systems and learn key optimization principles for building high-performance TinyML applications on open processor architectures.
RESOURCES:
drive.google.com/drive/u/2/folders/1wrmHH7-E2EFnB3hNlGg_Isgc69fLuaAD
SPEAKERS:
Niraj Dengale, Andes Technology Co. Ltd.
Frank, Yueh-Feng Lee, Andes Technology Co. Ltd.
This session was a part of the Software track at the RISC-V Developer Workshops held at RISC-V Summit Europe 2026 in Bologna, Italy.
WORKSHOP DESCRIPTION:
Hands-on TinyML Deployment on RISC-V: Building an Optimized Edge AI Pipeline - This hands-on workshop introduces a practical workflow for deploying TinyML applications on RISC-V edge platforms, demonstrating how developers can move from neural network models to efficient hardware-accelerated inference.
Participants will explore a complete TinyML deployment pipeline using an open RISC-V AI software stack. The session focuses on how neural network models designed for resource-constrained devices can be optimized and executed efficiently on RISC-V processors using optimized kernel libraries.
In collaboration with Edge Impulse (pending confirmation), attendees may explore how TinyML models can be developed and exported for embedded deployment using the Edge Impulse platform. If this collaboration is not available, the workshop will instead demonstrate a similar workflow using model preparation and optimization tools provided by Andes Technology.
The optimized models will then be deployed using neural network kernel libraries from Andes Technology, demonstrating how operator-level optimizations and efficient memory access patterns improve inference performance on RISC-V processors.
Example workloads may include representative TinyML models from MLPerf Tiny. These models illustrate how model architecture, kernel implementations, and RISC-V compute extensions such as vector or DSP instructions interact to deliver efficient edge AI inference.
Through this guided hands-on session, developers will gain practical experience deploying AI workloads on RISC-V systems and learn key optimization principles for building high-performance TinyML applications on open processor architectures.
RESOURCES:
drive.google.com/drive/u/2/folders/1wrmHH7-E2EFnB3hNlGg_Isgc69fLuaAD
