ILRI — Advanced Crop Model-Based Prediction for Farmers
International Livestock Research Institute (CGIAR) · AMFU · IMD
A weather forecast is not advice. The India Meteorological Department issues the forecast; Agromet Field Units translate it into agro-met bulletins for each district. That chain is sound science and it is national. What it cannot do is tell one farmer, standing in one field, what to do about the rain that is coming — for the crop they actually sowed, at the growth stage it has actually reached.
A volatile monsoon makes the generic bulletin obsolete
El Niño and La Niña have turned monsoon variability into the defining agricultural risk of the decade. An El Niño year can move the onset of the monsoon by weeks, break it mid-season, and compress the window in which a sowing, irrigation or spraying decision is still recoverable. In a stable year, a district bulletin covering many crops at many growth stages is a reasonable approximation. In a volatile one it is not — the same rainfall that rescues a crop at flowering will ruin one at harvest, and a bulletin written for a whole district cannot say which farmer is in which situation. The intelligence needed to draw that distinction already exists: IMD forecasts, the agro-met expertise accumulated in AMFU bulletins, crop calendars, and soil and rainfall records going back decades. It simply never reaches the farmer in a form they can act on, in a language they speak.
Turning the forecast into a bulletin written for one farmer
Kenpath is building an agro-meteorological advisory system with ILRI that generates bulletins at farmer scale. At its core is an advanced scientific crop model, built jointly by agricultural experts and AI experts, that simulates how a specific crop at a specific growth stage in a specific soil will respond to the weather IMD is forecasting. AMFU bulletin science, crop calendars and soil and rainfall records ground the model; AI does the work around it — generating the bulletin, and delivering it to the farmer in their own language by voice or by text. Every advisory carries its own audit trail: the forecast and field conditions it used, the decision it reached and the agronomic rationale behind it. AMFU experts validate the reasoning, not just the output.
The result
Agro-met intelligence that exists today only as a district bulletin becomes a crop-specific, stage-specific advisory written for one farmer — delivered in their own language by voice or text.
The crop model is built and validated by agricultural scientists working alongside AI engineers, so its recommendations carry agronomic authority rather than statistical correlation. The knowledge corpus and schema are designed so that adding a district, a crop or a language is a data operation, not a rebuild.
