Nordic Edge AI Lab

Nordic Edge AI Lab is a web-based platform for building and deploying AI models on Nordic Semiconductor's microcontrollers and SoCs. It lets you upload sensor data, choose a wake word, or configure a model, then train and download a ready-to-run inference model — no cloud dependency, no ML expertise needed. The platform supports two model types: Neuton for CPU-run edge AI on any Nordic SoC (time-series and sensor data) and LiteRT for SoCs with an Axon NPU (audio and advanced AI). Wake word models can be created without any training data.

The process is straightforward: upload data or pick a wake word, train automatically or configure architecture, then deploy. Built-in preprocessing tools handle windowing, feature extraction, and feature selection to shrink model size and boost accuracy. After training, you can analyze model quality with diagrams like feature importance and confusion matrices. Everything runs on-device — no raw data leaves the device.

Embedded engineers and IoT developers who want to add intelligence to low-power devices without deep ML expertise are the primary audience. Real-world uses include gesture recognition for smart home controls, wake word activation for voice interfaces, anomaly detection in industrial sensors, and health monitoring from wearables. It's a practical tool for making battery-powered devices smarter.