Cosh · Knowledge Engine

We spent years doing what
no one else has —
organising agricultural
knowledge at its roots.

Cosh is our proprietary knowledge engine. It structures what farmers need to know — every crop, every pest, every product, every relationship — into a living, multilingual knowledge graph that powers intelligent advisory at scale.

50,000+ Knowledge nodes
140,000+ Relationships mapped
12 Indian languages

Live preview of Cosh knowledge graph · Tap to explore →

Scroll to explore

The Problem

Agricultural knowledge exists. It just doesn't hold together.

Across India's farming landscape, the knowledge that helps a farmer protect a crop — which pest attacks which crop at which stage, what to apply, how much, in what language — has never been structured in one place. It exists in fragments: in experts' heads, in scattered documents, in advice that gets lost in translation.

Disconnected expertise

Agronomists know what works. But that knowledge lives in silos — unstructured, hard to transfer, and impossible to scale.

Advisory without context

Generic recommendations don't account for crop stage, local pest pressures, or the specific product a farmer can actually access.

Language as a barrier

India's farming communities speak dozens of languages. Knowledge in English doesn't reach the farmer who needs it in Kannada or Odia.

Products without meaning

Thousands of agri-input products exist — but understanding what each one does, when, and why, is buried in technical language that nobody translates.

What is Cosh

A knowledge graph built for the complexity of Indian agriculture.

Cosh doesn't store facts. It stores relationships. It knows that Root Knot Nematode affects Tomato during the vegetative stage, producing root galls — and it knows this in Hindi, Tamil, and Telugu. That is the difference between data and knowledge.

🌿

Agronomic Knowledge

Crops, growth stages, pests, diseases, symptoms, soil conditions, nutrition requirements, intervention timing — structured and interconnected.

📦

Product Knowledge

Agri-input products, dosages, manufacturers, application conditions, compatibility — connected to the agronomic context where they apply.

🔗

Semantic Relationships

Every connection carries meaning. IS A. AFFECTS. PRODUCES. IS APPLIED DURING. The graph doesn't just link — it understands.

Knowledge Architecture

Structured like a hypergraph. Designed for depth.

Cosh is built on a hypergraph model — similar in spirit to Neo4J, but purpose-built for agricultural knowledge. Cores hold the knowledge. Connects express the relationships. A single relationship can simultaneously express: which pest, at which stage, affects which crop, and produces which symptom. One statement. Complete meaning.



This architecture is what makes Cosh's knowledge not just storable — but queryable, traversable, and ready to power advisory logic that no flat database could approach.

Knowledge graph · simplified

Folder — Crop Stages

Organisational container

Core — "Vegetative"

Text node · 12 languages

Core — "Flowering"

Text node · 12 languages

Connect — Pest–Stage–Crop–Symptom

4-position hyperedge · Relationship types

Example relationship chain

Root Knot Nematode - Tomato
── IS ACTIVE DURING ──▶ Early Stage
── AFFECTS ──────────▶ Tomato
── PRODUCES ─────────▶ Root Galls

The Scale

Years of agronomic expertise, structured into one asset.

Built and curated by our internal agronomy team — not scraped, not auto-generated, not approximated. Every entry placed with purpose.

50,000+

Knowledge Nodes

Core data items covering crops, pests, nutrients, products, stages, and more

140,000+

Relationships

Semantic connections that give the knowledge its meaning and traversability

12

Indian Languages

Every knowledge node translated and expert-validated for farmer-facing use

2

Products Powered

RootsTalk and PesTest — with architecture ready for more

↳ These numbers grow continuously as our agronomy team extends coverage across more crops, regions, and product categories.

Multilingual by Design

Knowledge that speaks every farmer's language.

Translation is not a feature we added. It is part of the data model. Every knowledge node in Cosh exists simultaneously in English and in 12 Indian languages — curated with domain expertise, not just machine-translated.


When a farmer in Karnataka asks about a pest on their tomato crop, the advisory they receive is not a Google-translated approximation. It is expert-verified agricultural knowledge, in Kannada.

+ Assamese · Language registry is extensible — new languages added without code changes.

Hindi

हिन्दी

Tamil

தமிழ்

Telugu

తెలుగు

Kannada

ಕನ್ನಡ

Malayalam

മലയാളം

Marathi

मराठी

Bengali

বাংলা

Gujarati

ગુજરાતી

Punjabi

ਪੰਜਾਬੀ

Odia

ଓଡ଼ିଆ

Urdu

اردو

Gujarati

ગુજરાતી

Not Just Text

Knowledge has many forms. Cosh holds all of them.

Agricultural knowledge doesn't live only in words. A photograph of a pest. A video of correct spray technique. A document showing dosage tables. Cosh organises multimedia knowledge alongside text — tagged, language-aware, and ready to deliver.

🖼

Images

Pest identification photos, crop symptom images, product visuals — referenced in context.

🎬

Videos

Instructional content in multiple languages. A different video per language where needed.

🎵

Audio

Advisory content for farmers who prefer to listen — language-tagged and accessible.

📄

Documents

Technical sheets, label scans, application guides — structured and retrievable.

Explore Cosh

See the knowledge. Navigate the graph.

Every dot is a knowledge node. Every connection carries meaning. Pick a Core or a Connect, anchor on a crop or symptom, and watch the knowledge graph unfold. Hold a node, drag, rotate. Follow a connection — see where it leads.

Why Cosh is Different

What took years to build cannot be replicated overnight.

01

Human-curated, not machine-generated

Every entry in Cosh was placed by our agronomy team. Not scraped, not approximated. This is verified agricultural knowledge — and its quality shows in the advisory it produces.

02

Relationships, not records

Most agricultural databases store facts. Cosh stores the connections between them — the semantic relationships that let a system reason about what a farmer should do, not just what exists.

03

Multilingual at the core

Language is not a layer we apply on top. Every knowledge node exists natively in 12 Indian languages, with expert validation — not just automated translation.

04

Built to power products

Cosh already powers RootsTalk and PesTest. It is a tested, production-grade knowledge engine — not a prototype or a proof of concept.

05

Extensible by design

New crops, new languages, new product categories — the architecture grows without rebuilding. The same knowledge engine that powers today's products can power tomorrow's.

06

A proprietary asset, not infrastructure

Cosh is not a framework or a tool. It is a body of knowledge — structured, validated, and domain-specific. That makes it a genuine competitive moat, not just a technical advantage.

Talk to Us

Curious about what Cosh
could mean for your work?

We are exploring how Cosh's knowledge can create value beyond our own products — for researchers, agri-input companies, technology builders, and others working to improve outcomes for Indian farmers. If you see a possibility, so do we.

Start a Conversation

No forms. No sales pitch. Just a conversation.