Ambiq Micro has launched two AI runtime systems to boost AI performance and energy efficiency on edge platforms. The HeliosRT (Runtime) and HeliosAOT (Ahead-of-Time) are both optimised for the Ambiq Apollo Systems-on-Chip (SoCs) edge compute family.

The two systems are designed to address a growing need to deliver high performance within constrained power as AI workloads migrate to edge devices. Conventional AI frameworks frequently struggle on ultra-low-power platforms. This can make it difficult to deploy sophisticated AI models on battery-powered devices, such as wearables, hearables, IoT sensors, and industrial monitors.

HeliosRT and HeliosAOT are designed for edge AI across a wide range of applications, from digital health and smart homes to industrial automation.

HeliosRT is a performance-enhanced implementation of LiteRT, formerly TensorFlow Lite for Microcontrollers. It is tailored for energy-constrained environments.

It is compatible with existing TensorFlow workflows and introduces a number of improvements:

• Custom AI kernels optimized for Apollo510’s vector acceleration hardware
• Improved numeric support for audio and speech processing models
• Up to 3x gains in inference speed and power efficiency over standard LiteRT implementations

HeliosAOT introduces an ahead-of-time compiler that transforms TensorFlow Lite models directly into embedded C code for edge AI deployment. This offers runtime-level, or better, performance with additional benefits:

• 15–50% reduction in memory footprint versus traditional runtime-based deployments
• Granular memory control, enabling per-layer weight distribution across Apollo’s memory hierarchy
• Streamlined deployment, with direct integration of generated C code into embedded applications
• Greater flexibility for resource-constrained systems

“The intersection of developer experience and power efficiency is our north star,” said Carlos Morales, VP of AI at Ambiq. “HeliosRT and HeliosAOT are designed to integrate seamlessly with existing AI development pipelines while delivering the performance and efficiency gains that edge applications demand. We believe this is a major step forward in making sophisticated AI truly ubiquitous.”