Digitising agricultural research and field advisory

Context

Mahatma Phule Krishi Vidyapeeth is one of India’s principal agricultural universities, with a research and extension remit covering a large and agro-climatically varied region.

That variety is the difficulty. Guidance that is correct for a district average can be wrong for a specific plot with different soil, aspect and microclimate — and farmers notice quickly when advice does not match what they are seeing.

Why three products rather than one

The gap between agricultural research and field practice isn’t a single problem, it’s three. The farmer doesn’t have the data (weather and soil observations at plot resolution). The region doesn’t have the equipment visibility (where mechanisation actually sits and who can access it). And the research doesn’t reach the farmer in a form they can act on without an extension officer translating it.

Each of those needs a different kind of solution — a hardware product, a registry platform, and a decision-support model. Building all three together let the pieces share infrastructure, metadata and provenance discipline rather than emerging as three separate initiatives that would never quite line up.

What we built

Three connected pieces of work, each addressing a different part of the same gap between research and field.

The Smart Weather Station brought observation to the plot. Phule Sanman mapped the mechanisation actually present in the districts. OSDA turned research models into recommendations a farmer could act on without interpretation.

Weather at plot resolution

Regional weather stations are too coarse to drive plot-level recommendations in a district where aspect, elevation and soil change quickly. Deploying solar-powered single-pole stations at the plots that matter puts the data where the decision is made. The data is also archived at research-grade metadata — instrument identity, calibration, sampling cadence — so the same stream serves operational advisory and longitudinal research.

Mechanisation as a mapped network

Phule Sanman turns what was previously institutional estimation into a current registry — equipment owned, equipment available for custom hiring, repair services, spare suppliers. The platform is bilingual, low-bandwidth, and usable on the phones farmers already carry; institutional users get district-level gap analysis against the same underlying registry.

Research models that travel to the field

OSDA takes the research outputs — validated decision models, protocol packs, crop-stage-specific guidance — and renders them as direct, plain-language recommendations in the farmer’s language. Every recommendation traces back to the underlying research model and the inputs (soil, weather, crop stage, resources) that drove it, so an extension officer can explain the recommendation if asked.

Why it worked

The work was done with the institution’s own scientists rather than for them. Agronomic validity was set by the university; we were responsible for making it reach the field intact.

That separation of concerns — research correctness owned by the institution, delivery discipline owned by APx — kept both halves honest. The scientists did not have to become software designers to protect the integrity of their models; the engineering team did not have to become agronomists to get the delivery details right.

What generalises

The three-product pattern — plot-level sensing, mapped network of adjacent resources, model-based advisory in the user’s language — generalises to any research-to-field problem: horticulture, aquaculture, animal husbandry, forestry. The crops change; the architecture of delivery between institutional research and on-ground practice does not.