Start with your problem, not our pricing.
A short, time-boxed engagement on one brief of yours — a reformulation, an off-note to mask, a new SKU to ideate. We agree on the success criteria up front, run the engine on your ingredients with you, and you see an inspectable result before committing to anything bigger.
Mutual NDA first · forward-deployed · you see the result before any bigger commitment
How a pilot works
Four steps, scoped to one brief. No long integration, no committee.
Mutual NDA
Signed first — your formulations are protected as much as our method.
30-min scoping call
We agree the brief and 1–2 quantified success criteria, in writing.
Forward-deployed run
We run the engine on your ingredients with you in the room — not over the wall.
Inspectable readout
A short result you can interrogate input by input, with compound-level rationale.
Not a black box
The reason a pilot here is worth your time: you can audit every number it produces.
Glass-box scoring
Every per-pair score decomposes into its inputs — compound overlap (IDF-weighted), recipe co-occurrence (NPMI), perceptual/threshold weighting. Inspect it, don't take it on faith.
Paper-grounded
Built on 40,000+ peer-reviewed papers and a published methodology — not an opaque consumer-prediction model you have to trust.
Your data, your brief
Run on your ingredient list and your actual problem across 857 ingredients and 5,240 aroma compounds — not a generic catalog demo.
What a pilot can answer
Pick one brief. We agree the success criteria before we start.
Off-note masking
A compound-level read on what's driving an off-note, and directions to counter or cover it.
Criterion: 2–3 masking directions, with the molecules behind each.
Reformulation
Swap an ingredient (cost, supply, label) while holding the aroma profile, ranked by compound-set similarity.
Criterion: substitutes that hold the profile above an agreed similarity threshold.
New-SKU ideation
Novel, on-thesis pairings where the molecular and culinary signals agree — or productively disagree.
Criterion: N novel pairings your team hadn't considered for the brief.
The shape of it
Bring us a brief.
Tell us the problem and what a good result looks like. We'll send a one-page scope and a mutual NDA, and book the call.
Building software instead of running a brief? See the data & API →