Anya : Culturally Aware. Scientifically Precise.

Project-Theme-Image
a pie chart in meal generation showing food components
creation of meal plan according to geo-data
7-day custom plan being created with given geo-data accordingly
command center (prediction)
command center (fortification)
feature to add custom district location

Inspiration

India produces enough food to feed its entire population — yet 35.5% of children under 5 are stunted and 57% of women are anemic. We don't have a food crisis. We have a nutrition intelligence failure.

When we analyzed NFHS-5 data across India's 640+ districts, we found something striking: in Alirajpur, Madhya Pradesh, 79% of children are anemic — just 800km from Mumbai. These crises are predictable and preventable, but India's nutrition response is always too late, too generic, and culturally disconnected.

The data to solve this already exists — health surveys, food prices, weather patterns, PDS records — but it sits in silos, never connected, never predictive.

So we asked: What if one platform could predict which deficiencies will spike where, generate precision fortification strategies, and create culturally appropriate meal plans families will actually eat — all within their budget?

That became Anya — meaning "grace" in Sanskrit. Because every child deserves adequate nutrition, regardless of which district they were born in.

What it does

Anya is an end-to-end AI-powered nutrition intelligence platform built specifically for India. It operates as a three-stage pipeline:

1. PREDICT — Who's at risk?

Fuses data from NFHS-5, Agmarknet food prices, IMD weather, and PDS records to generate per-district, per-micronutrient risk scores (Iron, Vitamin A, Zinc, B12, Vitamin D)
Flags emerging hotspots before crises escalate
Covers India's top 10 most malnourished aspirational districts with real data for 1.6 crore citizens

2. FORTIFY — What nutrients are needed?

Generates cost-optimized fortification strategies per district using linear programming
Recommends the right fortification vehicle (PDS rice, edible oil, double fortified salt, mid-day meal flour) based on the district's specific deficiency profile
Incorporates bioavailability science — e.g., pairing iron with Vitamin C sources, separating iron and calcium across meals
Projects 3-month outcomes with estimated reduction in anemia, VAD, and zinc deficiency

3. NOURISH — Meals people will actually eat

Generates 7-day culturally aware meal plans for families based on:
- District and religion/cultural group (9 groups supported including Tribal Hindu, Muslim, Jain, Sikh, and more)
- Season and local food availability
- Monthly budget (as low as ₹2,500)
- Family composition (pregnant mothers, infants, elderly)
Uses only locally available, affordable ingredients including indigenous superfoods (ragi, moringa, drumstick leaves, wild amaranth)
Shows per-meal cost, nutrient coverage, and bioavailability tips
Includes a weekly food group breakdown and key nutrient sources specific to each cultural group

Our model calculates it costs just ₹847 per person to shift them out of nutritional risk through this pipeline.

How we built it

Frontend: React.js with a dark-themed command center dashboard, Mapbox for India district heatmaps, D3.js for charts and visualizations
Backend: Python with FastAPI serving the prediction, fortification, and meal generation engines
Risk Prediction: XGBoost model trained on real NFHS-5 district-level data (stunting, wasting, anemia rates) fused with food price indices, seasonal features, and poverty indicators
Fortification Optimizer: Linear programming (PuLP/SciPy) that maximizes deficiency reduction per rupee spent, subject to FSSAI safety limits and bioavailability interaction constraints
Meal Generator: Constraint satisfaction engine combining the Indian Food Composition Tables (ICMR), WHO/ICMR RDA values, cultural dietary rules database (9 groups), local market pricing, and PDS entitlement data
Data Sources: NFHS-5, Agmarknet, IMD, FSSAI fortification standards, ICMR Dietary Guidelines, Soil Health Card Portal

Everything runs on real government open data — no synthetic or mock datasets.

Challenges we ran into

Indian food is incredibly complex to model. A single dish like "dal" varies in ingredients, nutrition, and cost across every state. Building the cultural rules database for 9 religious/cultural groups with accurate dietary restrictions, pregnancy taboos, and fasting calendars was the hardest non-technical challenge
Data inconsistency across government sources. NFHS-5 has district-level data but Agmarknet food prices are at mandi (market) level — reconciling geographic granularity required careful mapping
Bioavailability modeling. Iron absorption varies 3-10x depending on what else is in the meal. Building the nutrient interaction matrix (iron + Vitamin C = good, iron + calcium = bad) and enforcing it in meal generation was scientifically demanding
Budget constraint optimization. Generating a meal plan that simultaneously meets nutrition targets, respects cultural rules, uses only locally available foods, AND fits a ₹2,500/month budget required solving a multi-constraint optimization problem in real-time
Handling "undefined" edge cases where certain districts had incomplete NFHS-5 data — we had to build fallback estimation logic

What we learned

Malnutrition in India is not a food problem — it's a data, delivery, and cultural sensitivity problem. The food exists. The government programs exist. What's missing is the intelligence layer connecting prediction to action
Cultural context is non-negotiable. A nutrition solution that ignores caste, religion, regional food habits, and pregnancy taboos will fail in India, no matter how technically impressive it is
Indigenous foods are nutritional powerhouses. Red ant chutney (iron-rich), ragi (calcium + iron), moringa (Vitamin A), wild amaranth — these traditional foods outperform expensive supplements, but have been systematically overlooked
Bioavailability matters as much as quantity. You can eat enough iron on paper but absorb almost none if paired with the wrong foods. This invisible factor is a major driver of India's anemia crisis
Government open data is a goldmine — NFHS-5, Agmarknet, Soil Health Cards, IMD — the infrastructure for intelligent nutrition planning exists, it just needs to be connected

What's next for Anya

Expand to all 640+ districts with full NFHS-5 coverage and real-time Agmarknet price feeds
Integrate with POSHAN Tracker — the government's existing Anganwadi monitoring platform — so Anya's predictions and meal plans reach 1.4 million Anganwadi workers directly
Add an Anganwadi worker mobile app with offline-first architecture, voice-based UI in 12+ Indian languages, and camera-based MUAC (mid-upper arm circumference) measurement for instant child malnutrition screening
Pilot in Nandurbar, Maharashtra — partner with the district ICDS office to deploy the fortification strategy and meal plans across 50 Anganwadi centers serving ~2,000 children
WhatsApp-based meal plan delivery — send personalized weekly meal plans to mothers via voice messages in their local language, reaching the 500 million Indians already on WhatsApp
Real-time food price integration — dynamically adjust meal plans when local ingredient prices spike, automatically substituting affordable alternatives
Open-source the platform so other countries with similar malnutrition challenges can adapt Anya to their context