Day 1
Conception
The 1,000-day window opens. Maternal nutrition and antenatal care already shape the child's growth trajectory.
ALU E-Lab · Think Tank · 2026
L.E.A.P
Leaders Executing African Progress
We turn the health data Rwanda already collects into an early-warning signal against child stunting — inside the first 1,000 days, while prevention still works.
01 — Mission
To reduce child stunting in Rwanda's most affected communities by turning the health and household data already being collected into an early-warning signal — equipping Community Health Workers and health authorities to identify and support at-risk children during the critical first 1,000 days, before growth damage becomes irreversible.
02 — The Problem
More than one in four children under five in Rwanda is stunted.
33%
children stunted in 2020
DHS 2020
27%
children stunted in 2025
DHS 2025
19%
national target
still out of reach
Stunting — chronic impaired growth caused by prolonged malnutrition in the first 1,000 days of life — damages cognition, education, and lifelong earning potential during a child's most critical developmental window. After that window closes, the harm is irreversible. The burden is geographically uneven, and it persists because agriculture produces calorie-rich but nutrient-poor diets while healthcare manages the consequences — neither sector owning the solution. Data is collected every month; nothing turns it into timely, targeted action.
The burden is geographically uneven — stunting under five, DHS 2025
- Nyarugenge (urban Kigali)8.7%
- National target19%
- National average27%
Urban Kigali records the lowest rates; rural Northern and Western districts remain far higher — and children born to mothers without formal education face the greatest risk.
The first 1,000 days — conception to age two
Day 270
Birth
Growth monitoring begins. CHWs start collecting the records our model reads — weight, feeding, household data.
Day 450
First foods
Complementary feeding begins. Nutrient-poor staples like cassava and beans fill stomachs but starve growth.
Day 730
Growth faltering
Risk compounds quietly. A child who is falling behind still just 'seems small' — no signal says act now.
Day 1,000
The window closes
Around the second birthday, stunting becomes permanent. Everything before this line is prevention; everything after is management.
03 — Why It Persists
We asked why five times.
Why 01
Programs treat what you can see
Food supplements and antenatal care address the visible symptoms, not the cause. Stunting doesn't look obvious — a child just seems small — so families and clinics catch it late, while health, agriculture, and sanitation actors work in separate lanes.
Why 02
The right foods aren't on the plate — and nobody owns the problem
Households eat what fills you up — beans, cassava — not what a growing child needs. Growth check-ups happen inconsistently, and stunting falls through the cracks between departments because no single sector is responsible for it.
Why 03
Farmers grow what's safe; health workers are stretched thin
Smallholders plant staple crops because they're reliable and affordable, not nutritious. CHWs cover too many households with too few resources, so the household data they collect doesn't always lead to action.
Why 04
Poverty limits choices; budgets chase visible wins
Many families can't afford diverse foods. Governments and donors fund what shows results quickly — but preventing stunting takes years to appear in the data, so prevention keeps losing to urgent-looking problems.
The Root
Data accumulates — but never becomes a signal
CHWs and health authorities gather growth, feeding, and household records every month at national scale. Nothing converts that data into a prediction of which children are about to become stunted. Information piles up without ever triggering timely, targeted action.
04 — The Solution
The only solution that predicts which children will become stunted — before it happens.
Rwanda already runs systems that locate children who are stunted; by design, they are reactive. Our machine learning model reads the same underlying data — growth monitoring records, household socioeconomic category, maternal health indicators, feeding data — and generates a predictive risk score per child during the first 1,000 days, so intervention happens before growth faltering becomes irreversible, not after.
01
Data already collected
Growth monitoring · household records · maternal health · feeding data
02
Interpretable model
A classifier trained on existing records — validated against real outcomes
03
Traffic-light flag
Low / medium / high risk, with the reason stated in plain language
lowmedhigh
04
Targeted action
Home visit · referral · nutrition counseling — within days, not years
Pillar 01
No new infrastructure
It rides on data and CHW workflows already in place. No new devices, no retraining, no new household spending.
Pillar 02
A flag, not a dashboard
A CHW with no data-science background sees a simple low/medium/high signal — and why it was raised.
Pillar 03
Validated against outcomes
Predicted risk is compared with actual growth outcomes months later, so accuracy claims are earned, not assumed.
05 — The Prototype
CHW view · Musanze District
Child record 0417
14 months · household cat. 2
Why flagged
Flagged primarily due to feeding pattern and two missed growth checks in the last quarter.
Recommended action
Home visit within 7 days · nutrition counseling referral
12 low4 med1 high
A machine learning model that predicts individual-level stunting risk before it occurs, built on NHIC's existing real-time health data. One model, two layers: a CHW-facing risk flag per child, and an aggregate risk dashboard for health authorities.
Precision
% of high-risk flags confirmed as actual stunting cases
Recall
% of stunting cases correctly predicted in advance
Lead time
months of warning before clinical detection
Response rate
% of flags leading to a visit or referral within 7 days
Must-have
- Trained classifier outputting a risk score per child
- Interpretable low/medium/high flag — not a raw probability
- Validation against known outcomes before any pilot claim
Should-have
- CHW-facing flag screen mockup
- District-level aggregate risk view for NHIC
- Feature-importance: why was this child flagged?
Could-have
- Semi-live connection to a sample NHIC dataset
- Feedback loop: predictions checked against outcomes over time
- Kinyarwanda interface for CHW-facing output
Won't-have
- Data collection or cleaning pipelines — NHIC owns these
- A production mobile app
- Physical interventions — those are what a flag triggers
06 — The Journey
Six challenges. One term. Documented, not described.
Challenge 01 / 06
Meet the Think Tank
Six students, one conviction: Africa's hardest problems deserve builders, not bystanders. In our launch video we introduce Leaders Executing African Progress — who we are, why we formed, and why we chose to stand in front of child stunting.
Challenge 02 / 06
Help Lab
Before building anything, we helped. We walked with two families through the barriers between them and healthcare — transport, cost, information — and saw first-hand how distance and income decide who gets care in time.
Challenge 03 / 06
Expert Interview
We sat down with the Rwanda Biomedical Centre to pressure-test our thesis. The interview confirmed the gap our research kept finding: the data to catch stunting early already exists — what's missing is the signal that turns it into action.
Challenge 04 / 06
The Grant Pitch
We pitched The First 1,000 Days Link for $10,000 — a budget mapped to model development, validation with a partner health center, and the NHIC partnership pathway. Every dollar goes to prediction, because prevention is where the leverage is.
Challenge 05 / 06
The Perception Challenge
A documentary on the Mursi of Ethiopia's Omo Valley — a people the world flattened into a stereotype. Making it taught us the discipline our venture depends on: look again, past what a community 'seems' to be, to what the evidence says.
Challenge 06 / 06
Web Design
You're looking at it. This site is the sixth challenge: our journey, told as a night signal — built with Next.js and GSAP, designed and coded by the think tank it represents.
You're looking at it.
Next.jsGSAPTypeScriptDesigned & coded by L.E.A.P
07 — The Think Tank
Six people standing between the data and the damage.
Pacifique Mico
Team Lead · Data & Research
Leads the think tank — and turns DHS and EICV datasets into evidence, and evidence into the numbers this site is built on.
Bellamy Dan Biramahire
ML Engineering
Owns the model — from feature design to validation — and keeps the venture honest about what the data can and cannot claim.
Bruno Dushimiyimana
Health Systems & Partnerships
Maps the road from prototype to NHIC partnership, navigating data governance and the institutions that scale it.
Latonia Igihozo
Community Research
Led our interviews with mothers, CHWs, and cooperative leaders — the voices that shaped the empathy map.
Boneya Abagudo
Strategy & Operations
Keeps six people, five challenges, and one workbook moving in the same direction, on time.
Noella Akimana
Communications & Design
Shapes how L.E.A.P speaks — from the grant pitch narrative to the visual language of this site.