Alif Semiconductor released the DK-B1 development kit for its new Balletto microcontroller (MCU), claimed as the first Bluetooth Low Energy (BLE) wireless MCU to feature hardware optimization for AI/ML workloads. It enables developers and manufacturers to implement sophisticated AI/ML functions such as speech recognition, AI noise cancellation, keyword spotting, audio beamforming, and sensor fusion in wearable and other types of space-constrained devices.
The Balletto B1 wireless MCU combines a 160-MHz Arm Cortex-M55 CPU core with Helium vector processing extension, an Arm Ethos-U55 NPU, a BLE and 802.15.4 radio, and digital and analog capabilities in BGA or CSP packaging.
The company also annouced that it will support the ExecuTorch framework recently introduced by PyTorch, which enables on-device inference capabilities across mobile and edge devices. This will provide users of Alif’s Ensemble and Balletto devices with a more seamless path to design, optimize, train, and deploy machine-learning models directly on endpoint devices. Enabling data and decisions to be processed immediately on-device, it allows for secure personalization, reduces power consumption, and increases the flexibility of on-device AI use-cases.
Alif is also collaborating with Arm to address hardware acceleration of transformer-based models in microcontroller devices. This would open the door to the creation of language models and other advanced use-cases to deploy on edge devices using ExecuTorch.
