A physics-informed AI for the alcohol industry.

SensoryOps was started on the observation that the alcohol industry already has the data, the sensors, and the PLC stacks – but lacks a layer that turns shelf-level taste signals into chemical setpoints a production line can actually run against.

The closest comparables – Gastrograph AI on the consumer side, PhysicsX on the physics-informed-AI side – solve half the problem each. SensoryOps sits at the intersection: a physics-constrained predictive layer that takes a flavour target as input, simulates what this specific vessel will produce, and pushes the optimised setpoints to the line.

The technical bet is that physics-informed neural networks, trained on this vessel's geometry and this grain bill's enzyme profile, generalise past any point the on-line sensors have seen – because the physics constraints hold. Validation across a 60-minute mash shows 1.27% L² error on sugar concentration; you can see the chart on the home page.