Smoke Signal
Smoked chipotle and paprika, built for a slow, rounded heat rather than a sharp one.
The model tested 400+ chipotle-to-paprika ratios; our chef locked in 3:1 for balance without scorch.
We use an AI flavor-pairing model to test thousands of spice ratios in minutes — then hand every blend to a working chef before it ships. Nothing reaches your kitchen untasted.
Model disclosure: blend ratios are generated by a proprietary AI flavor-pairing model trained on culinary combination data. The model proposes candidates only — it does not approve products. A human chef tastes, adjusts, and signs off on every blend before it's listed for sale.
Three blends, each run through hundreds of AI-generated ratio candidates and narrowed down by hand.
Smoked chipotle and paprika, built for a slow, rounded heat rather than a sharp one.
The model tested 400+ chipotle-to-paprika ratios; our chef locked in 3:1 for balance without scorch.
Sumac, lemon peel, and coriander for a sharp, herbal lift on fish, chicken, and greens.
The model flagged sumac-to-lemon ratios most tasters preferred; the chef added coriander to round the acidity out.
Cardamom, cumin, and black pepper — a slow, warm base built for braises and roasted roots.
Highest-rated candidate from the model's warm-spice cluster; the chef cut the cinnamon by a third.
The AI model proposes hundreds of ingredient ratio candidates from a culinary pairing dataset.
A working chef tastes the top candidates and adjusts salt, heat, and acid balance by hand.
The finalized blend is re-tested on a full production batch for consistency before packing.
Every listing states what was AI-assisted and what a human decided. No exceptions.
A short, real walkthrough of a blend moving from AI-generated ratio candidates to a chef's final adjustments.