Ambiq has introduced two innovative edge AI runtime solutions, HeliosRT and HeliosAOT, specifically designed for their Apollo SoCs to elevate AI performance and energy efficiency within the realm of edge computing.
HeliosRT represents a power-optimized iteration of LiteRT (TensorFlow Lite for Microcontrollers), offering remarkable enhancements in inference speed and power efficiency, with up to 3x improvements. HeliosAOT serves as an ahead-of-time compiler that transforms TensorFlow Lite models into embedded C code, resulting in reduced memory usage by 15–50% and enhanced deployment adaptability.
Ambiq has launched HeliosRT and HeliosAOT, cutting-edge edge AI solutions tailored for Apollo SoCs to boost AI performance and energy efficiency in edge computing.
Both these cutting-edge solutions tackle the challenges associated with deploying AI on ultra-low-power devices such as wearables, IoT sensors, and industrial monitors. Leveraging Ambiq’s patented SPOT technology, these tools bring substantial improvements in power consumption for edge AI applications.
Carlos Morales, VP of AI at Ambiq, emphasizes, “The fusion of developer experience and power efficiency guides our development. HeliosRT and HeliosAOT are meticulously crafted to seamlessly integrate into existing AI development pipelines while delivering the required performance and efficiency for edge applications. This marks a significant stride towards making sophisticated AI ubiquitous.”
HeliosRT is presently available in beta, with a full release slated for Q3 2025, while HeliosAOT is in technical preview for selected partners, with broader availability scheduled for Q4 2025.
These tools seamlessly blend into existing AI development workflows and are backed by comprehensive documentation, examples, and engineering support.
Related
AI/ML | Ambiq | edge AI | Edge Inference | IoT
