Technology
How SMART BAKE AI works
The system connects three levels: the experimental production that generates the data, the computational processing that turns it into models, and the application that returns predictions on a human scale.
The chain
From dough to model and back
01
Experimental production
Each batch is prepared as a discrete research unit. The input variables are recorded — composition, dosages, conditions — together with the observed results.
02
Computational processing
Cleaning, normalisation and transformation of the data; development of input features; training and comparative evaluation of models.
03
Experimental application
The selected models are integrated into a web environment that accepts parameters and returns experimental predictions with an uncertainty range.
Inputs and outputs
What goes in and what comes out
Input parameters
- Flour type and protein content
- Hydration — water on flour
- Fat, sugar, salt, yeast
- Fermentation temperature and time
- Baking temperature and time
Experimental predictions
- Specific volume
- Crumb moisture
- Crumb hardness
- Crust lightness
- Estimated shelf life
What sets it apart
A research tool, not a black box
It states its limits
When a parameter falls outside the experimental range, the system says so explicitly instead of returning a seemingly valid value.
It explains the prediction
Sensitivity analysis shows which parameter contributes what — interpretability, not just a number.
It gives a range, not certainty
Every prediction is accompanied by an experimental uncertainty range.
What it is not. The system is an experimental research prototype and is not a finished commercial product. The predictions do not replace trial production and do not constitute a guarantee of results.
Outlook
Where it may lead
The new knowledge, the organised base of experimental data and the models create the conditions for the business to support evidence-based decision-making in the design and modification of recipes and to direct subsequent experimental trials more effectively. In the future, this use may contribute to reducing the number of unproductive trials, the consumption of raw materials and the development time for new products.