Technology
Two engines. One honest week.
µLayer builds every dish the way a home kitchen does, layer by layer. µSolve chooses the dishes for each meal and works out every gram against 43 nutrient targets from ICMR-NIN 2020.1
Nutrient targets
43
From ICMR-NIN 2020, each with a floor, a goal and a safe upper limit.
Foods in the table
577
Indian food composition data, with a few staples from USDA, cooking losses applied.
Dishes, built in five layers
449
269 Andhra dishes in Telugu and English, and 180 from across India.
µSolve 1.0 · the plan engine
It works out the numbers. It doesn’t guess them.
Chatbots write a menu and estimate its nutrition as they go. µSolve starts from your targets and safe limits instead, and searches real plates for the grams that come closest. The same food always counts the same.
It decides
- Which plate each meal gets. A yes or no for every candidate, exactly one yes per meal.
y ∈ {0, 1} - How many grams of each ingredient. Any amount inside the recipe’s range, and only on the plate it chose.
lo·y ≤ g ≤ hi·y
It must keep
- Energy within 5% of the day’s target
- Protein at least 97% of the goal
- Every daily upper limit
- The caps for the client’s conditions
- Added salt under 5 g, or 3 g
- Portion limits for every meal
- No dish twice in a day
- Your Must and Never rules
It works to make small
- Every shortfall below a goal, weighted by the nutrient. Below the floor costs three times as much.
Σ w·(d₁ + 3·d₂) - Every overrun of added sugar, saturated fat and sodium.
- Repetition of dishes and foods from earlier days.
Where the targets, food values and cooking losses come from: the science.
Every meal, a real plate
It never invents a dish.
µSolve chooses only from µLayer’s plates: dishes that fit how you eat and what you avoid, put together the way your cuisine serves them. Each day it draws a fresh set, leaning towards what the day still needs.
- 449dishes in two cuisine packs269 Andhra and 180 pan-Indian, each one built in layers.
- 5checks for this clientDiet type, allergies by every name, foods you exclude, gluten or lactose, and your Never rules.
- 30candidate lunches a dayPlates drawn by what the day still needs and what the week has not had: 22 breakfasts, 30 lunches, 14 snacks, 26 dinners.
- 1plate per mealChosen and sized by µSolve.
Seven days, balanced as a week
A short day is made up the next day.
ICMR-NIN 2020 sets vitamin and mineral needs over your usual intake, not a single day. So µSolve plans the days in order, and when one falls short, the next day’s target rises by 60% of the gap. Energy, fat and salt are held day by day.
- 820of 1,000Mon
- 1,090of 1,108Tue
- 1,010of 1,054Wed
- 930of 1,048Thu
- 1,080of 1,090Fri
- 1,040of 1,042Sat
- 1,000of 1,018Sun
- Planned that day, mg
- That day’s target
- Week: 996 mg a day
µLayer 1.0 · the recipe engine
Every dish, built in five layers.
A recipe book gives fixed amounts. µLayer builds each dish the way a home kitchen does, layer by layer: a protein core, the grain, the vegetables and the helpers that season them, each with its own range. So when µSolve changes 45 g of dal to 52 g, your method still says 52 g.
- 1
Protein coreDals and pulses, peanuts and sesame, curd, milk, paneer, eggs, meat and fish.
A vegetarian lunchMoong dal 30 g, curd 150 g12.4 g proteinA prawn dinnerPrawns 130 g18.5 g protein - 2
SupportGrains and nuts with more protein: wheat atta and rava, bajra, little millet, almonds.
A vegetarian lunchNot in this mealA prawn dinnerWhole-wheat flour 80 g8.5 g protein - 3
OuterThe rest of the grain: rice, jowar, ragi, maize and the other millets.
A vegetarian lunchRice 85 g6.7 g proteinA prawn dinnerNot in this meal - 4
FibreVegetables, greens, tubers and fruit.
A vegetarian lunchAmaranth 165 g, pumpkin 80 g, brinjal 70 g, drumstick 50 g8.4 g proteinA prawn dinnerGongura 90 g1.7 g protein - 5
HelpersOil, the tempering, souring, aromatics and measured salt.
A vegetarian lunchOnion 45 g, mint 25 g, tamarind 11 g, oil 10 g, the tempering, salt 1.7 g5.5 g proteinA prawn dinnerOnion 40 g, tomato 40 g, ginger and garlic 7 g, oil 5 g, spices, salt 1 g1.9 g protein
Inside a dish
Thotakura pappu, layer by layer.
The dal is the protein core, the amaranth the fibre, and the oil and tempering are helpers. Each flexible part stays within the range a real cook would use. Cooking losses are worked out for every food, and the added salt is counted in every dish.
తోటకూర పప్పు
Thotakura pappu · amaranth with toor dal · Andhra · 40 min
µSolve sizes
Toor dal kandi pappuProtein core
30–60 g45 g
Amaranth leaves thotakuraFibre
80–128 g104 g
Groundnut oil palli nuneHelpers
3–6 g4.5 g
Helpers µLayer fixes
- Tomato
- 40 g
- Green chilli
- 3 g
- Garlic
- 3 g
- Turmeric
- 0.3 g
- Iodised salt
- 0.8 g
Taalimpu, the tempering
- Mustard seeds
- 0.8 g
- Cumin seeds
- 0.5 g
- Dry red chilli
- 0.6 g
- Curry leaves
- 1 g
- Asafoetida
- 0.05 g
Then it writes
- Rinse 45 g toor dal and cook it with 3 parts water until very soft; pressure-cook for 3–4 whistles. Mash well.
- Cook the leaves with the turmeric, slit green chilli, 40 g chopped tomato and a splash of water, about 5 minutes. Stir in the dal.
- Heat 4.5 g groundnut oil and add the mustard, cumin, red chilli, curry leaves and asafoetida with the garlic. Pour this taalimpu over the pappu.
After cooking losses, one serving
- 243 kcal
- 14.2 g protein
- 393 mg calcium
- 7.1 mg iron
Inside a meal
A plate has rules too.
µLayer knows how Andhra homes serve dishes together: what goes beside a pesarattu, when curd finishes a rice meal, and the pairings nobody would make. µSolve only ever chooses between plates that follow these rules.
An Andhra lunch
- అన్నంRiceThe base
- పప్పుPappu, pulusu or curryThe main
- కూరKuraA vegetable beside it
- పచ్చడిPachadiOn the days you set
- పెరుగుPeruguOften, curd to finish
Pairing rules it never breaks
- One main vegetable per plate
- Never two tamarind dishes together
- No fresh pachadi beside a meat curry
- A tiffin comes with its own chutney: pesarattu with allam, idli with kobbari
How often each lunch is drawn
- Rice plate 64%
- Roti plate 16%
- One-pot rice 20%
In every diet plan
Every plan is built by both engines.
What changes from plan to plan is who checks the week, and how often it is planned again.
Self-guided
µSolve and µLayer plan your week from what you tell us, for healthy adults.
See Self-guidedDietitian-reviewed
A registered dietitian checks and edits what the engines wrote, with your health, medicines and blood work in view.
See Dietitian-reviewedPremium
The engines plan four different weeks at once, and your dietitian reviews the whole month, with a nutrient certificate.
See PremiumCoaching, coming soon
The engines plan each new week from your check-in, and your dietitian approves every change.
See CoachingWhere AI fits
No chatbot picks a food or a gram.
µSolve and µLayer are arithmetic over published tables, so every number in your plan traces back to a food and a rule. Where food alone can’t reach a target, such as vitamin D, your plan says so instead of rounding up.