Working on something new

Dec 7, 2025

A few months back, I started training ML models for time series classification with Nordic’s Thingy:53. It was my first end-to-end embedded ML project in years. Apparently it’s called Edge AI now. Tools like Edge Impulse make it easy to train and deploy models to embedded platforms. But here’s what you quickly realize: you need a lot of data to make your model accurate enough for the real world. And if you’re not working with a demo application that comes with a dataset, collecting that data yourself is a real pain. You need large volumes of it, and you need to label it carefully. Edge Impulse has great interfaces for importing data, but they don’t provide dedicated collection tools. They created a demo mobile app for the Thingy:53, but it lacked data visualization, which is crucial for verifying you’re capturing the right events. The app was mostly just a gateway between the device and the web application. So for the last few months, I’ve been building the ideal data collection app for my own Edge AI projects. I’m a few weeks away from launching it on the app stores. I’ve got some cool ideas of where to take it, and I’m planning on building in public as much as possible. Follow along with me through more posts here and on X.