Biohacker Formulation Intelligence
Food Formulation Optimization Software
Choose ingredients, set nutrition targets and weight limits, and a deterministic optimizer calculates candidate proportions that meet them. It chooses among the ingredients you select. It does not invent new ones.
Free sandbox with synthetic demo data. Not laboratory validated. Nothing you enter is saved.
Loading the synthetic ingredient catalog…
Worked example with synthetic data
A 40 g serving mixes a synthetic protein blend, a synthetic oat powder and 4 g of a synthetic cocoa component. The targets are at least 20 g of protein and at most 1 g of sugars, at the lowest ingredient cost. The optimizer returns 22.29 g of protein blend, 13.71 g of oat powder and 4 g of cocoa component: exactly 20 g of protein, 0.54 g of sugars and USD 0.5486 of fictional ingredient cost. Rounded to 0.5 g steps it becomes 22.5, 13.5 and 4 g, and every target still holds.
How the optimization works
For each ingredient the unknown is its weight in one serving. The weights must add up to the serving mass, stay inside the limits you set, and make each nutrient total meet its target.
Σ xᵢ = M
Lᵢ ≤ xᵢ ≤ Uᵢ
Nⱼ = Σ xᵢ · qᵢⱼ ÷ 100
Nⱼ ≥ aⱼ ; Nⱼ ≤ bⱼ
C = Σ xᵢ · pᵢ ÷ 1000
Among all weights that satisfy those hard constraints, a linear program finds the one with the lowest ingredient cost, or the highest protein within a cost budget. The solver is HiGHS.
Every answer is then audited again from the raw numbers, without trusting the solver status. Rounded weights are re-checked too. If any check fails, no formula is shown.
A second candidate appears only when it is genuinely different and answers a clearly stated alternative objective.
Method version: formulation-lp/1.0.0
What this does and does not do
- It models a dry mix with no processing losses. It does not predict flavor, texture, solubility, shelf life, safety or bioavailability.
- Results are point estimates from the data supplied. They are not laboratory, regulatory or manufacturing validation, and they support no nutrition or health claim.
- A missing value is unknown, never zero. If a target needs a value that is missing, the calculation is blocked.
- Cost covers ingredients only. This sandbox uses fictional prices, never supplier quotes.
- A real project needs measured data for your actual ingredients and a food technologist to review the result before any prototype.
Data and sources
The sandbox catalog is synthetic and exists only to show the method. A production catalog will use public reference data, such as USDA FoodData Central (CC0), and supplier specifications, each with its source, basis and review recorded.
Formulation questions answered
Does the tool invent recipes with AI?
No. It does not use a language model for any number. A linear-programming solver chooses weights among the ingredients you select, and an independent audit checks the result.
Can I trust the nutrient numbers?
In this sandbox they are synthetic on purpose. In a real project they would come from documented sources and still be estimates until a laboratory measures the finished product.
Is my formula stored or shared?
No. The sandbox stores nothing, uses no cookies for it and sends no formula to analytics or advertising tools. Only the request you submit yourself reaches us.
Why did I get no formula?
Either the targets cannot be met by the selected ingredients and limits, or a needed value is missing. The tool says which, and never shows a made-up closest recipe.
How do I work on my own ingredients?
Request a feasibility review. Our team can scope a project with your specifications. A private workbench for confidential formulas is planned and is not available yet.