First virtual sensor developed

We just taught our existing sensors to be a sensor they're not - using a 30KB ML model that lives on our chip. Smaller BOM, lower cost, higher reliability, zero new parts.

At Pickpad we use hardware to fix challenges software can't solve alone, building the missing infrastructure layer at the intersection of restaurants, retail, customers, and delivery. But our internal strategy remains the same: less hardware, more compute.

Every component we don't add is a failure point eliminated, a supplier dependency avoided, and a permanent saving on every unit we ship. That's why our PCB has only 42 components - and we keep making it less, while adding new features.

Here's a recent example. We needed a new sensing capability on Pickpad. Normal route - a dedicated sensor, ~5 new parts, added cost forever. Our route - a micro ML model under 30KB, running on our existing chip, taking data from sensors we already have and simulating an entirely different sensor in software. New capability, zero new parts, works perfectly. Patent pending.