DuKGest™

DuKGest™ is the nutrition engine behind PATO’s food logging. When a member logs a meal — by photo, voice or text — DuKGest™ is what turns that into accurate macros. Its defining characteristic is what it doesn’t do: it never lets a language model produce a nutrition number.

Most AI nutrition tools ask a language model what’s in a meal, and the model answers. It sounds right, it’s formatted confidently, and it’s frequently wrong — because a language model generating “340 calories” is producing plausible text, not looking anything up.

DuKGest™ splits the job in two:

1. Recognition — identify what the food is, and roughly how much. This is genuinely what AI is good at.
2. Valuation — retrieve the nutrition values for that food from a catalogue of 11K+ verified foods, and calculate the totals on a server.

The AI never touches step 2. It can be wrong about what it’s looking at, and a member can correct that. It cannot be wrong about what a boiled egg contains, because it was never asked.

The distinction the whole category turns on.

Nutrition tracking has one failure mode that kills it: members stop trusting the numbers. Once someone spots an obviously wrong calorie count, the whole feature becomes decoration — and any coaching built on that data becomes guesswork.

It also matters for what you can responsibly say. A coach can work from grounded numbers. Nobody should be adjusting a member’s intake based on figures a language model invented.

DuKGest™ runs behind every meal a member logs, whether they photograph it, say it, or type it. Members can correct anything, and the correction sticks. Their pantry improves accuracy further, because the system stops guessing which yoghurt they buy.