🚢 couldn't be happier to ship my favourite hardware hack yet: we built Tandem, a lightweight AI-driven urban bike assistant at this year's @HackTheNorth!
here's the story behind it (full deets at https://t.co/fcRE4X5AjB):
I’m still in a bit of disbelief that we managed to get it all working (with 90%+ accuracy!)—massive thanks to @HackTheNorth for organizing <3
ly all @_pranavnt@JasonYuan869 @freemanxjiang, thank you for making this a weekend to remember!
That evening, we’d go on a few dozen test rides all across Waterloo (most of which weren’t on bike lanes), sensing 90% of cars behind us and storing everything in real-time :)
peek our stupid camera setup for demos lol
Over the next 14 hours, we’d collectively have an ungodly amount of Awake chocolate as our team built night vision, open-source location + incident tracking for cities, passing & blind spot alerts, and make it all work offline and with nothing more than a mobile power bank
By noon that day, we finished off our modelling & APIs, and headed out of E7 for our first test run around the parking lot of @HackTheNorth (and somehow, it worked!)
After lugging our hardware halfway across the continent (ty @_pranavnt) and a bike 100mi to Waterloo, we mounted a depth-sensing OakD camera to the back of my mountain bike and built a box around a tiny Raspberry Pi for on-device ML & a HUD
Cities are unscalable right now—they're sprawling, asphalt-filled messes hostile to anyone outside of a car.
If we want more liveable cities, we need meaningful alternatives to this.
The US & Canada don't want to build bike infrastructure, so we decided that we would instead.