I’m very excited to announce my next chapter as the co-founder of Edge Impulse. Together with Zach Shelby we want to make machines smarter, by enabling developers to create the next generation of intelligent device solutions through embedded Machine Learning.
We believe that in the near future machine learning will become a standard development tool for solving problems that rule-based programming approaches currently cannot easily solve. This might range from detecting animal behavior (‘has this cow been acting abnormally?’) and detecting acoustic events (‘was there glass breaking?’) to detecting fault states in machines earlier (‘is this washing machine at risk of failure?’).
We also believe that you can only make sense of this data by processing and analyzing at the very edge. Devices already capture lots of data, but 99% of sensor data is discarded today due to cost, bandwidth or power constraints. If you need to send raw sensor data to the cloud you’ll drain batteries very quickly, and run up a high bill from your telco.
But… deploying ML on end-devices is hard. It requires embedded developers to capture clean and correct data of devices, signal processing experts to extract meaningful features, data scientists to build a correct machine learning model, and then more specialized skills to transform a trained model into something that can run in kilobytes of RAM. We’ve founded Edge Impulse to make this easier. Our online tooling will allow developers to correctly collect data, create meaningful data sets, quickly spin up models, and generate open source code for rapid product iteration.
This is just the start of the journey, but stay tuned! If you’re a developer that is interested in applying ML to real devices, and want to provide early feedback on our solution, drop me a message!
For some more background, also read Zach’s announcement blog post which goes into more detail, and also has prettier pictures!