Bengaluru firm launches micro language model for climate-smart farming

Aksara promises to provide farmers with the right kind of information, hyper-tuned to local conditions and crops, so that they can make climate-smart decisions.

The existing large language model, which served as the base model for Aksara, is fed with more generic data, which are biased towards the Global North, said Krishna Kumar, founder and CEO of Cropin.

“What we have done with Aksara is that we fine-tuned contextual data on nine kinds of crops — paddy, wheat, maize, sorghum, barley, cotton, sugarcane, soybean and millets — with particular reference to countries in the Indian subcontinent,” added Kumar.

Interview: BJP candidate and ‘Mysuru Maharaja’ Yaduveer Wadiyar wants to script his own destiny

AI-based knowledge domain

The knowledge domain of the model is specific to the agricultural best practices, including climate-smart agricultural practices and regenerative agricultural practices, said Kumar.

“AI has the potential to transform agriculture, but challenges like access to large-scale structured data, expertise, and storage/compute infrastructure limit its adoption. Open-source projects like Aksara are crucial for wider AI use and speeding up innovation,” said Praveen Pankajakshan, head of Cropin AI Labs.

Eventually, Cropin plans to build a version of Aksara, which will not only monitor but also alert the farmers about possible interventions to mitigate the problem at hand.

“Anybody can download the model. With more people downloading and building on it, it is possible in the near future for Aksara to process real-time information and suggest interventions that are beneficial,” said Kumar.

According to him, for now, Aksara is programmed to provide responses on focus crops and countries. The model is trained on a database containing information from seed sowing to harvesting, covering every phenological stage of the crop growth cycle and different aspects like crop health management, soil management, disease control, and others, said Kumar.

More geographies and crops will be added later and by leveraging multiple modalities of data, it will be possible to get intelligent climate data, added Pankajakshan.

(Disclaimer: The headline, subheads, and intro of this report along with the photos may have been reworked by South First. The rest of the content is from a syndicated feed, and has been edited for style.)