EdgeWave – Edge AI Platform for Smart Sensor Applications

EdgeWave is semify's edge AI platform, currently in development. It brings AI inference directly to the sensor, so data is processed locally in real time instead of being sent to the cloud. It combines hardware-accelerated AI with semify's expertise in digital design, verification and RISC-V systems.

From model to silicon: the full AI flow
We handle the complete path from a trained neural network to working hardware, from one source:

• Model development and training in TensorFlow

• Model optimization, quantization and pruning for efficient hardware

• Mapping the model onto a dedicated hardware accelerator

• RTL design and full verification, including UVM

• Implementation on FPGA, or integration as an IP block inside your ASIC

FPGA or ASIC: your choice
EdgeWave can run on FPGA for fast prototyping, flexible products and smaller volumes. It can also be integrated into an ASIC for the lowest power, smallest area and high-volume production. We support you on both paths, and on the move from FPGA prototype to ASIC.

Current development
EdgeWave is being developed together with an industrial partner, starting with an advanced acoustic sensing application. This lead use case drives the architecture toward demanding requirements: continuous signal processing in real time, tight power budgets and inference on the device itself. Because the platform is designed for these constraints from the start, it can carry over to other sensor types with little extra work.

Designed for

• Real-time inference at the edge with deterministic, low latency

• Low power use, suited to battery-powered and always-on devices

• Data that stays on the device, which improves privacy and reduces bandwidth

• One partner for the whole path, from AI model to verified hardware

• Custom RISC-V instructions that can be added for AI workloads

Outlook: beyond acoustics
EdgeWave is designed as a general platform for intelligent sensor processing. Future applications include:

• Vibration and condition monitoring for predictive maintenance

• Industrial sensing and anomaly detection

• Environmental and gas sensing

• Motion, gesture and presence detection

• Medical and wearable sensing

Early partners wanted
We're looking for partners and customers who want to bring AI into their sensor products. Joining at this stage lets you shape the platform around your application, whether as a joint development, a pilot project or through licensing and IP integration.