Context
Our client supports 160+ wheat millers and aims to ensure every batch of flour is fortified with essential micronutrients, especially iron.
Iron fortification is a proven, cost-effective way to combat iron deficiency and is mandated in over 100 countries.
But testing whether a batch contains the right amount of iron remains a challenge. Existing methods are often slow, expensive, and prone to human error.
Our client was particularly interested in using Artificial Intelligence to improve the iron spot test—a widely used but qualitative method (present/absent)—and turn it into a fast, reliable, and quantitative (numerical) approach for real-time decision-making.
Devised a method to create synthetic datasets for training as the cost of collecting ground-truth data is very high.
Built a comprehensive training dataset to support model development.
Trained a machine learning model to estimate added iron in wheat flour from the images of the iron spot tests.
Launched a simple mobile responsive webapp to help users analyze iron content easily in the field.
