
Never a food photo
The leading calorie apps depend on photographing your plate after you eat. Rasoi tells you what to serve before the meal and asks for one tap after. Leaving out the camera is the whole idea.
Confirming a family dinner: one tap.
A household nutrition operating system for Indian homes. What each person should eat, how much to cook and serve from one shared dish, and what to buy. When real life gets in the way, it re-plans. Never a food photo.
Working prototype · closed beta in preparation
The household nutrition
operating system
Every Indian home answers four questions, every day: what should we eat, how much, what do we need to buy, and what do we already have. Rasoi answers all four, every day, for every person at the table, from one shared dish.
Usually one person carries all four questions in their head. Today's apps answer one or two of them and hand the rest back. Rasoi is built for the household, not the lone dieter: one pot of dal, and a different serving for each person who eats from it.



The problem
Existing products answer one or two of these questions and leave the rest to the person who already carries the load.
Different people, different goals: a vegetarian and a peanut allergy at the same table.
One pot of dal, but a teenager, a parent losing weight and a grandparent need different servings.
The weekly list is guesswork. Something always runs out and something always goes to waste.
Nobody knows what's left in the kitchen until it's time to cook.
Why now
India's plate got easier, not healthier. For sixty years, convenience decided what the country grows, mills, rations and cooks. The bill is arriving as a public-health crisis.
of India's total disease burden is attributed to unhealthy diets.1
adults are living with diabetes, and another 136 million have prediabetes.2
of women aged 15–49 are anaemic, as are 67% of children aged 6–59 months.3
of daily energy comes from carbohydrates, mostly white rice, milled grains and sugar. Protein: about 12%.4
It is a double burden. More than one in three children under five is stunted, while 24% of women and 23% of men are overweight or obese, up from 21% and 19% in 2015–16.3 More than a third of adults have high blood pressure.2
How it happened
Rice and wheat got the high-yield seeds and the prices. Land under millets shrank by about 60%, to 14.7 million hectares.5
The state buys 45–70% of India's wheat and rice, but only 1–3% of its millets.6
Millet flour keeps only 20–30 days, so it is harder to mill, ship and stock.6
Cheap rice and wheat arrive through the ration system. Millets are now 6–7% of the staple diet.6
Oil makes food quicker and tastier. Use per person is up 84% in rural India since 2004–05.8
62% of energy now comes from carbohydrates and 12% from protein, and saturated fat is over the healthy limit in all but four states.4
Millets fell from 32.9 kg to 4.2 kg per person a year between 1962 and 2010. Wheat almost doubled, from 27 kg to 52 kg.
−87%millets eaten per person, 1962 to 2010.
+93%wheat eaten per person over the same years.
| Grain | 1962 | 2010 |
|---|---|---|
| Millets | 32.9 | 4.2 |
| Wheat | 27 | 52 |
~3×Overall oil use per person has nearly tripled since 2001, to about 23.5 kg a year.9
56%of the edible oil India uses is imported.8
| Area | 2004–05 | 2022–23 |
|---|---|---|
| Rural | 5.76 | 10.58 |
| Urban | 7.92 | 11.78 |
What it costs
The cost shows up on the plate. People eating the most carbohydrate had 30% higher odds of newly diagnosed type 2 diabetes than those eating the least. Swapping some of it for plant or dairy protein was linked to lower risk.
With oil, it is the kind of fat more than the amount: on average total fat stays within national guidelines, but saturated fat is above the limit for metabolic health in all but four states, and healthier fats stay low.4 In April 2025 the Prime Minister backed a call to cut cooking oil by 10% to fight obesity.9
These are associations from a large national survey, not proof of cause.
Where Rasoi comes in
Rasoi doesn't ask a household to give up what's easy. It makes the balanced plate the easy one: protein, vegetables and grains measured out per person from one shared pot, oil given in millilitres instead of poured by habit, and the shopping list already written.
See how it works
Sources
“Tell the household what to cook and how much before the meal. Then ask for a single tap afterwards. Never a photograph of the plate.”
How it works
Calorie apps ask you to record what you ate after the fact. Rasoi works the other way round, and the next plan knows what really happened.








Then the loop starts again, and the next plan knows what really happened.
Why it's different
“An AI tells you what to eat” is crowded, and general assistants are already there. Three product laws make Rasoi something else, and we don't bend them.

The leading calorie apps depend on photographing your plate after you eat. Rasoi tells you what to serve before the meal and asks for one tap after. Leaving out the camera is the whole idea.
Confirming a family dinner: one tap.

The household is the user. Rasoi says what belongs on each plate and how much, served from one shared dish. Dishes are suggestions, not orders. That's how dietitians actually coach.
~150 g cooked grain · ~100 g dal · ~120 g vegetable · 1 tsp oil

Cooking changes nutrients, and by how much is uncertain. Rasoi gives ranges with a confidence label. It never gives falsely exact percentages or dresses up an estimate as a measurement.
“Steaming generally keeps more vitamin C than long boiling” (confidence: medium)
Everyone in the home with their diet, allergies and goals, editable as life changes.
Exact servings in grams from a shared dish, sized to each person's targets.
Ingredient amounts, oil in ml, salt in grams, the method, and nutrient notes with a confidence label.
As planned, less, more, swapped, skipped. A whole family in a single tap.
Shelves grouped by aisle, quick restock, and short check-ins when stock is uncertain.
The week's plan minus what's at home, as a list you can take to any store.
Ate out? Skipped lunch? The rest of the week adjusts, within each person's upper limits.
An excluded food never reaches the planner for that person, so it can't end up on their plate.
Choose how many days before a dish comes back. Everyday staples like roti, rice and sambar don't count as repeats.
Each pot and each plate shows the salt the recipe adds. How much to use stays your call; the plan doesn't change.
Which meals got an answer, and how many took a single tap. No calorie scoreboard.
Off until you turn them on. Scheduled on the phone, and the lock screen never shows a name or a dish.
Where we are
The engine is built and the full loop works end to end. The hard part came first: nutrition arithmetic, safety rules and the meal optimiser, each with automated tests. Then came the app around it.
Every value records its source, version and confidence, and is checked against hand-worked examples.
Excluded foods are removed before planning starts. The release test allows zero violations.
Plans the whole household at once, and stops with a stated reason if no safe plan exists.
93 plates from 61 recipes, from poha and dosa to ragi mudde and fish curry. No dish comes back sooner than the household chooses, and each meal carries its share of the day instead of one oversized plate.
Every core task has a tap budget. Optional meal reminders stay on the phone. The Android beta build is set up; signed builds and device testing come next.
The loop closes: a deviation re-plans what's left of the day and keeps the rest of the week. The shopping list asks for a packet of garam masala, not each of its spices by the gram.
Privacy controls were built before the beta, not bolted on after. Deleting an account removes the sign-in too, not just the data.
A registered dietitian has reviewed the nutrient targets, recipes and plates over seven rounds. Allergen declarations go to a food-safety reviewer next. Required before any real household relies on the plans.
The first beta starts with 28 dishes. Their cooked weights and servings will be measured at the stove, not estimated.
Measuring the one number that matters.
Walkthroughs, October 2026. A family of three (one peanut allergy, two vegetarians) went through consent → 7-day plan → meal taps → pantry → grocery basket → unplanned meal → re-plan → data export. A second household, set up from scratch: an adult over sixty, a heavy-activity adult and a member allergic to milk and gluten, all vegetarian. In both, no excluded food reached anyone; in the second, all 70 servings checked were vegetarian.
What households told us
We asked thirty households what's hardest. The answers were mixed, which argues for one product across meals, quantities and groceries, not another single-purpose app.
What share of meals are confirmed with one tap on an existing plan, rather than fresh manual entry?
Household interviews, September 2026, n = 30. An exploratory sample, not a population estimate. Observed usability sessions come next.






Roadmap & business model
Prove the habit first, then build the business on it. Each phase has a test it must pass before the next begins. No spend on growth until households keep coming back.
Specs, household interviews, food data, targets, safety rules, optimiser.
API, mobile app, plan → confirm → pantry → grocery → re-plan.
100–300 households. Passes if one-tap confirmation holds and people are still using it at week 8.
Subscriptions, family plan, a web app for household admin, support.
Ordering through grocery retailers and ONDC; a dashboard for dietitians.
A mode for clinicians, after a regulatory assessment.
How Rasoi makes money
One profile, a basic dashboard and a manual list. The way in.
Annual billing from day one. The family plan costs more because it replaces several tools.
Commission on baskets the plan already produces. Always disclosed, and always ranked below nutrition and safety.
Professionals manage client households on the same engine.
No advertising. We never sell health profiles. Either would break the trust the product depends on.
Built to be trusted
An AI that explains the plan, but never makes up the numbers. These rules are written into the design, and automated tests check the app follows them.
Language models explain decisions and understand requests. Amounts, portions and targets come from a tested calculation engine, never from a model.
A food a person must avoid is removed before planning starts. It isn't a preference the optimiser can trade away. An intolerance is excluded exactly like an allergy, traces included.
Each nutrient value records where it came from, which version and how confident we are. Every plan keeps an audit record.
Any sponsored grocery item is labelled and always ranked below suitability and safety. We don't experiment on that rule.
Consent controls processing. Export or delete your household's data at any time. The app never uses the camera.
A missed meal isn't piled onto dinner. The rest of the week is re-planned within each person's upper limits.

Early access & investment
We're preparing a closed beta with urban Indian households, and talking to investors who back products that take trust seriously.
If that's you, or if your family wants to try Rasoi first, write to us.