KissanAI has launched Dhenu2, its latest agricultural AI platform to address the challenges of agriculture. Initially focused on India, the platform will gradually expand globally, offering solutions to farmers, policymakers and agribusinesses.

Powered by the Llama 3 architecture, Dhenu2 is available in three versions: 8B, 3B and 1B models, meeting different deployment needs. The platform is based on 1.5 million real and synthetic instructions, covering more than 4,000 agricultural categories and topics.

Insights were drawn from conversations with over 100,000 farmers, providing real-world data for optimized decision-making.

The Dhenu development platform was also introduced, allowing developers to create custom farming solutions using Dhenu2. With six years of agricultural data curation and synthetic data pipelines for efficiency, the platform is open source, providing flexibility to developers and organizations.

Dhenu2 is designed to provide practical information, thereby helping stakeholders in the agriculture sector make informed decisions.

KissanAI was launched in March 2023 by Pratik Desai. The platform primarily targets Indian farmers, providing them with region-specific agricultural knowledge, often difficult to access due to language barriers and limited digital literacy.

The company was recently accepted into the NVIDIA Inception program. “We are excited to be part of NVIDIA Inception and leverage their resources to make our AI capabilities faster, more accurate and more efficient for farmers,” said Pratik Desai, CEO of KissanAI. “This partnership will accelerate our mission to boost agricultural productivity and sustainability. »

Last year, KissanAI released Dhenu 1.0, an agricultural LLM for Indian agriculture built on top of Llama 2. This model includes English, Hindi and Hinglish, catering to farmers’ language preferences. It is trained on large datasets and processes 300,000 instructions in both languages, improving support for farmer queries.