Listen to this story
|
KissanAI today announced the launch of Dhenu 1.0, a groundbreaking Agriculture Large Language Model. Tailored specifically for Indian agricultural practices, this bilingual model comprehends English, Hindi, and Hinglish queries, a notable feature catering directly to farmers’ linguistic needs.
Founder Pratik Desai disclosed that Dhenu 1.0 was meticulously trained on extensive, high-quality datasets intricately focused on Indian agricultural practices. The model’s uniqueness lies in its bilingual nature, adeptly processing 300,000 instruction sets in both English and Hindi. This innovation enables Dhenu 1.0 to comprehensively support English, Hindi, and Hinglish queries, catering directly to farmers’ linguistic needs.
KissanAI also partnered with Sarvam AI and NimbleBox.ai for this project. The collaboration with the Sarvam AI team, renowned for its pioneering work in Indic language AI, brought bilingual capabilities to the fore. NimbleBox.ai’s exceptional AI API platform played a pivotal role in expediting the data curation process, a crucial component of the development of Dhenu 1.0.
Desai also expressed gratitude to Microsoft for Startups’ continuous support and TTeknium1’s Open Hermes 2.5, acknowledging its cost-efficient handling of substantial data volumes.
Initial human evaluations have shown promising results, but the team plans rigorous testing before deployment due to its high-impact nature. Desai emphasised that this unveiling marks only the beginning of KissanAI’s efforts to revolutionise agriculture, hinting at future collaborations and innovations.
KissanAI, founded by Desai, is the brainchild of the son of an Indian farmer based in the US who wanted to do something for Indian farmers back home. “We’ve been working in this space for some time, we have enough data in agriculture, and we are working with farmers very closely,” he said.
Desai has a PhD from Wright State University and has been building AI/ML applications for agriculture for quite some time. Previously, he developed an automated labelling mechanism using generative AI. “Using Stable Diffusion within the first month of its release, we created nearly 20,000 stock images,” Desai said.
Desai is constantly partnering with agricultural universities in the country to keep the data up-to-date. The platform has a voice interface that supports nine Indic languages, including Gujarati, Marathi, Tamil, Telugu, Kannada, Malayalam, Punjabi, Bangla, and Hindi. Pretty soon, two more languages – Assamese and Odia – will be added to the list.