NYUS

From the founder

India's Digital Health Stack Is Missing Food Data

India has built digital rails for health IDs, records and claims. The data layer for what people actually eat, home-cooked and unlabelled, is still missing.

By Shaan, founder of NYUS ·

India has spent the last several years building genuinely ambitious digital health infrastructure. In May 2026 the National Health Authority reported that the Ayushman Bharat Digital Mission (ABDM) had crossed 90 crore Ayushman Bharat Health Accounts (ABHAs), up from 14.7 crore in 2021. Each ABHA is a 14-digit health identity that lets a person link and share their health records with consent across providers, and the same mission runs the National Health Claims Exchange for insurance claims. Whatever one thinks of the pace or the privacy trade-offs, the architecture is real: India is building a national data layer for health.

There is one layer underneath all of it that nobody has clearly been asked to build: food.

Every health intervention eventually asks what you ate

Diabetes management, cardiac risk reduction, maternal nutrition, weight management, even routine wellness coaching: each of them eventually needs to know what a person eats. In markets where most food is packaged, digital-health tools have leaned on labels. In the US, for example, the Department of Agriculture's FoodData Central combines laboratory composition data with a Branded Foods dataset built from product label information supplied through a public-private partnership. A consumer app, an insurer wellness programme or a dietetics tool can lean on that kind of infrastructure: scan the product, look up the numbers, log the meal.

That model does not transfer well to how most Indians eat. A katori of dal, two rotis, a bowl of sambar, a plate of biryani made to one household's recipe: none of it has a barcode or a label.

What India's reference data covers, and what it does not

India does have a serious reference dataset. The Indian Food Composition Tables, 2017, published by the ICMR-National Institute of Nutrition, provide data on 151 food components for 528 key foods. It is a valuable, carefully built resource. But by design it catalogues ingredients: except for eggs, the data are for foods in their raw form. It does not try to describe a cooked dal, a restaurant curry or a home-made sabzi, and it cannot capture the variation that matters most in home cooking: how much oil went in, how a recipe drifts from region to region, and portion sizes judged by eye rather than on a kitchen scale.

Anyone who has tried to log an Indian meal knows the problem. Even something as plain as a roti is not one number, as we found when comparing roti sizes. Any AI system, health app or clinical tool that tries to reason about an Indian patient's actual diet runs into this gap almost immediately, and as far as we can tell from the public record, no one has closed it at national-infrastructure scale.

Why this is a policy question, not an app problem

As India's digital-health stack matures, more of it will touch diet directly: non-communicable disease (NCD) risk scoring, diabetes and hypertension programmes, insurer wellness incentives, even AI triage tools that ask about eating habits as part of an assessment. The stakes are large. The ICMR-INDIAB national study, published in The Lancet Diabetes & Endocrinology in 2023, estimated the weighted prevalence of diabetes among Indian adults at 11.4% and of hypertension at 35.5%, and concluded that metabolic NCDs in India are considerably more common than previously estimated. Diet is a central part of how both conditions are managed.

Every one of those systems will eventually need to answer a deceptively simple question: what did this person actually eat, and how much protein, carbohydrate and fat was in it? And every one of them will hit the same wall. The underlying data does not exist in a form a machine can use, because it was never built for a country where so much of what people eat is cooked at home.

What building one small version taught us

We have spent real engineering effort on a narrow version of this problem while building NYUS, an AI fitness and nutrition coach: mapping 8,900+ foods, from Indian staples to global foods, to usable macro data, and building a logging flow that starts from a katori and a roti rather than a barcode. Barcode scanning exists for packaged food, but it is the minority case. We mention this not as a pitch but because it is the closest thing we have to direct evidence of how large the gap is.

Even at the scale of a single app, mapping Indian home food to nutrition data took sustained, deliberate work. It did not fall out of any existing public dataset, and every dish entry is still an estimate of a typical preparation rather than a measurement of the plate in front of you. That is why practical workarounds such as estimating portions with your hand matter so much for Indian users. If closing the gap for one consumer app takes that much effort, closing it for a national digital-health stack is a considerably bigger undertaking, and it is not currently anyone's clearly assigned job.

Who should own the food-data layer?

That is the policy question worth asking. As India builds out ABDM, health IDs and interoperable clinical records, who owns the food-data layer underneath them?

We do not have a confident answer to who should own it. What seems clear is that the question is underdiscussed relative to its stakes. India has done the hard, unglamorous work of building digital rails for identity, health records and insurance claims. The next unglamorous infrastructure problem, one that touches nearly every future AI-driven health application built for this market, is food. Right now, it is still a blank spot on the map.

In the meantime, individuals tracking their own diet can still get useful, approximate numbers; our guide to calories and macros in common Indian foods is a starting point. This article is not medical advice. If you are managing diabetes, hypertension or another condition, work out your diet with a doctor or a qualified dietitian.

NYUS is in early access

An AI fitness and diet coach built for Indian food: a weekly check-in that adjusts your calories, macros and step goal from what you log, 8,900+ foods with macros, in English, Hindi and Hinglish. No ads, no data sold.

Join the early-access list

Frequently asked questions

What is an ABHA number?

ABHA (Ayushman Bharat Health Account) is a 14-digit digital health identity under the Ayushman Bharat Digital Mission. It lets a person link, access and share their health records with consent; the National Health Authority reported more than 90 crore ABHAs in May 2026.

Do the Indian Food Composition Tables include cooked dishes like dal or roti?

No. The Indian Food Composition Tables, 2017 from ICMR-NIN cover 151 components for 528 key foods, and except for eggs the data are for raw foods. Cooked and home-made dishes have to be estimated from their ingredients.

Why is it hard to track nutrition for Indian home-cooked food?

Home-cooked dishes have no barcode or label, and oil, recipe and portion size vary from household to household. Any app or tool can only give approximate values based on a typical preparation.

Sources

  1. ABDM crossed 90 crore ABHAs (May 2026), up from 14.7 crore in 2021; ABHA is a 14-digit consent-based health identity; ABDM includes the National Health Claims Exchange
  2. USDA FoodData Central includes a Branded Foods data type built from label data through a public-private partnership
  3. IFCT 2017 (ICMR-NIN) covers 151 food components for 528 key foods; except for eggs, data are for raw foods
  4. ICMR-INDIAB (Lancet Diabetes & Endocrinology, 2023): weighted prevalence of diabetes 11.4% and hypertension 35.5% among Indian adults; metabolic NCDs considerably higher than previously estimated