Pratik Desai, founder of four-year-old agricultural artificial intelligence startup Kissan AI, last month announced Dhenu 1.0, a large-scale agricultural language model with seven billion parameters.

The model is named after the divine cow goddess Kamadhenu, whom Hinduism describes as the mother of all cows. The model is trained on high-quality conversational data that specifically focuses on agricultural practices in India. Dhenu is an LLM in Agriculture specially designed to help Indian farmers in agriculture-related matters.

It is designed bilingually and trained on 300,000 sets of instructions in English and Hindi.

“Our goal was to develop an AI/ML application that can be used by farmers. Throughout our journey we have experimented with building different tools and have been expanding our knowledge base,” says Desai.

In March last year, the company launched KissanAI, a chat platform for farmers in 10 Indian languages. Farmers could speak in their own language and get answers to questions related to agricultural inputs, policy packages for various crops, fertilizers and pesticides.

Desai realized that for his solution to scale and reach farmers, it had to be cheap, even though GPU usage costs were expensive for him.

“Our goal was to build a small model based on more than 355,000 agriculture conversations we had had with more than 100,000 farmers, and voice or text datasets we had collected on our KissanAI platform,” adds Desai.

Some of the material comes from agricultural information brochures or images from brochures. Thus, the company used optical character recognition to digitize the text and engaged in tagging and curating the data. They worked with three agricultural universities to get the terminology right.