Running AI on small devices is challenging. A new software platform aims to simplify deployment and speed up product development.
Ambiq Micro has launched heliaCORE, a software platform designed to speed up AI deployment on resource-constrained edge devices powered by Apollo system-on-chips (SoCs). The software combines optimized AI kernels, operator acceleration, and validation tools to improve the performance and efficiency of AI workloads while reducing development effort.
Built on the Arm CMSIS-NN framework, heliaCORE accelerates not only neural network kernels but also key inference operations such as activations, padding, reductions, reshaping, and data movement. By optimizing these functions, the platform aims to improve real-world AI execution on edge hardware while remaining compatible with existing embedded AI development workflows.
The software includes more than 200 accelerated operators optimized for Apollo devices and supports MVE and DSP acceleration across both current and previous generations of Apollo SoCs. Developers can integrate heliaCORE into existing projects with minimal workflow changes.
To assess deployment readiness, Ambiq validated heliaCORE across more than 40 AI models, covering 53 operator types, 247 unique operators, and 963 operator instances. According to the company, this represents roughly 12 times more operator instances than MLPerf Tiny, providing broader coverage of real-world AI workloads.
heliaCORE is the latest addition to Ambiq’s HELIA Edge AI ecosystem, joining heliaRT and heliaAOT. Together, these technologies form the company’s software stack for developing and deploying AI-enabled products on Apollo-based hardware.
The software is available through CMake, CMSIS-Pack, Zephyr, and neuralSPOT-X integration flows, enabling developers to incorporate it into existing toolchains and accelerate product development.
“The future of AI will be defined by intelligent devices operating at the edge, but moving AI from experimentation to real-world deployment remains one of the industry’s biggest challenges,” said Dr. Adam Page, Head of AI at Ambiq. “heliaCORE is a foundational building block of the HELIA ecosystem, providing the software infrastructure developers need to deploy edge AI faster and with greater confidence.”
Click here for the original announcement.
