Video here: https://www.youtube.com/watch?v=8XiP2-DPo6g

This is what I’ve been working on full-time as a solo founder in my startup, Sentinel Devices, for just over four years. High-level: we collect data from the “digital brains” of industrial equipment, then use this data to build a baseline model of equipment behavior and monitor for anomalies – signs in the digital data indicating maybe that a sensor or valve is failing, or that some kind of cyber-physical attack is ongoing. The unique part about our approach is that we are laser-focused on doing everything totally offline – that means all data collection, storage, processing, AI training, and decision-making happen exclusively on embedded industrial hardware. There are no external servers, data is never sent anywhere, and our models are not even pre-trained – our devices self-train and self-develop models in the field, based only on data they’ve seen. By removing the need for cloud connectivity, we’re building something that scales infinitely to as many machines as you’d like – while not introducing any of the issues that come with an always-on external network connection.

I thought this would be relevant given the (very) recent events around always-on internet connections and updates bricking critical infrastructure. We’ve developed OTAware specifically for critical infrastructure – if there’s no internet connection, attackers can’t get in (or at least it’s a lot harder) and you won’t be impacted by your infrastructure making random updates. Definitely welcome any thoughts or discussion from industrial folks around use cases you see, or obstacles to deployment you can think of. And if anyone would like to try it out, it does require physical hardware, but we’ll happily talk with you about what a test deployment can look like!