AI Developer Kits

heliaCORE™, heliaRT™, and heliaAOT™ are the underlying kernel and runtime layers, built on Ambiq’s SPOT® platform, that actually execute AI inference on Apollo silicon. An ADK sits above them — it’s the ready-made workspace (tasks, datasets, training config, demos) that lets you build a specific application, like heart-rate detection, and then deploy it through whichever runtime fits your product.

Ambiq features a variety of ADKs with new models and features constantly being added. A few of the ADKs include, heartKIT™ for cardiac signal processing (denoising, ECG/PPG segmentation, rhythm and beat classification) aimed at fitness wearables and remote patient monitoring; sleepKIT™ for on-device sleep analytics (sustained sleep/inactivity detection, sleep staging, and apnea/hypopnea detection); soundKIT™ for voice AI (speech enhancement, keyword spotting, voice activity detection, and speaker identification) for wearables, earbuds, and smart home devices; physioKIT™ for physiological signal processing in ambulatory health monitoring and wearable technology offering  tools to process, analyze, and synthesize signals commonly captured by wearable sensors, including ECG, PPG, RSP, and IMU data; and compressionKIT™ which helps teams compress continuous sensor waveforms before they become a memory, radio, battery, or cloud-ingestion problem, designed for wearable and edge signals such as PPG, ECG, IMU, and future time-series modalities where preserving waveform utility matters.

Both. Each ADK ships with pre-trained demo models so you can get a working prototype in hours, but the YAML-driven workflow lets you swap in your own datasets, model architectures, and training settings without refactoring the pipeline — you’re not locked into Ambiq’s defaults.

It depends on your product stage: TFLM gives the most portable baseline, heliaRT™ is a drop-in, backward-compatible speedup that suits a model still being iterated on, and heliaAOT™ compiles ahead-of-time for the smallest, most power-efficient build once your model is locked down. 

ADKs are purpose-built for battery-powered, always-on products: heartKIT™ targets fitness wearables, wellness devices, and remote patient monitoring; sleepKIT™ targets sleep-tracking wearables and remote monitoring; and soundKIT™ targets wearables, earbuds, and smart home devices that need on-device voice AI. All three are optimized for Ambiq’s ultra-low-power Apollo SoCs.