In 2022, Cropin launched what it claimed was the ‘world’s first cloud for agriculture’. Now, it is adding generative AI into the mix.
The Bengaluru-based agritech firm just launched Cropin Sage, which it claims is the ‘world’s first real-time agri-intelligence platform’.
Powered by Google Gemini, Cropin Sage enables users to make informed decisions based on historical, present, and future data on cultivation practices, crops, irrigation, climate, and soil.
To provide accurate data at such a scale, Cropin has partitioned the global map into 5×5 km grids, assigning a unique ID to each. Users can query Sage regarding crop cultivation feasibility within specific grids. Sage then provides data visualisation outputs in response to these inquiries.
The data is then processed by Sage, which sits inside the Cropin cloud, into different grid sizes and aggregated at different temporal frequencies, including yearly, seasonal, monthly, weekly, and daily, based on customer requirements.
(Cropin Sage)
Gemini Queries Large Datasets
Sage leverages Cropin’s advanced crop models, which analyse crop performance in detail. It also leverages a smart climate model, which integrates data spanning the last 40 years and current conditions and forecasts weather for the coming year.
“We integrate Gemini with our climate models, crop models, and proprietary knowledge graphs. These elements are fused together to provide comprehensive solutions,” Krishna Kumar, chief executive officer at Cropin, told AIM.
At its core, Sage makes sense of the enormous datasets spanning over terabytes of data. Each dataset could easily be over 100 gigabytes in size, depending on the country and the availability of historical data, besides other datasets such as temperature, climate, and even socioeconomic data.
In fact, Sage uses Gemini to query these large datasets and generate responses in a consumable manner within seconds. The startup’s proprietary data spans over 350 crops and 10,000 varieties in 103 countries.
(Cropin Sage)
Sage does not present the data in a text format but leverages a visualisation tool to transform complex datasets into intuitive graphs, charts, and interactive displays that enhance understanding and decision-making.
To do so, Sage leverages the Gemini Flash 1.5 model to convert user queries to SQL queries, which allows it to generate user-friendly, grid-based data in a visually appealing platform.
It also leverages Google Kubernetes Engine (GKE) services to scale its operations in real-time as demand increases, processing massive volumes of data.
According to Kumar, Sage will initially provide information on 13 crops expanding to 15 by the end of the year. These crops, which include corn, paddy, soy, wheat, onion, potato, sugarcane, cocoa, coffee, and cotton among others, accounts for 80% of total global production.
Moreover, users can seek data on these 13 crops irrespective of where they are. “They could be anywhere, any country, but if they seek data on these crops, they will get the data,” Kumar emphasised.
Who Benefits from Sage?
Rajesh Jalan, the CTO and head of engineering at Cropin, who was also present during the conversation with Kumar, told AIM that the model is highly accurate.
Hallucinations remain a persistent issue with LLMs. In fact, upon Gemini’s release, it generated controversial responses that led Google CEO Sundar Pichai to deem them ‘unacceptable’.
However, according to Jalan, it does not matter here because Sage provides all the data as per the query. “The customers can see all the data and if the data is wrong, the customer will be able to figure it out.”
So far, the company has nearly 250 B2B customers worldwide and has digitised over 30 million acres of farmlands, positively impacting over 7 million farmers worldwide.
Cropin Sage will benefit CPG players, seed manufacturers, food processors, multilateral organisations, financial institutions and governments, according to Kumar.
Global food systems today are encountering substantial challenges, exacerbated by various adverse factors that significantly hinder farmers’ capacity to meet food production demands.
This issue is exemplified by the cocoa crisis, which has deeply affected chocolate manufacturers. Cocoa prices have surged by 400% within a year, posing significant affordability challenges for many in the chocolate industry.
“Consider an enterprise in the food and agribusiness sector with a supply chain spanning multiple countries like PepsiCo, for instance. Due to an unstable supply chain caused by changing climate changes, if they want to expand the production of potatoes to a new country, they can leverage Sage to predict suitable locations for sourcing and investments,” Kumar said.
Sage can also benefit multiple governments, especially in countries which lack good agricultural data.
“For instance, we’re collaborating with the Kenyan government on corn production challenges, analysing grid-level data to understand climate impacts and crop trends, addressing food security concerns comprehensively,” he added.
Akṣara – Cropin Micro Agri Model
Earlier this year, Cropin announced the launch of ‘akṣara’, the sector’s first purpose-built open-source micro language model (µ-LM) for climate-smart agriculture.
Built on top of Mistral’s foundational models, the startup aims to make agri-knowledge accessible to everyone in the ecosystem. “We trained it with data for five countries and nine crops and now we are embedding the model with Sage,” Kumar revealed.
While Sage is a data-intensive platform, akṣara is a knowledge platform, and according to Jalan, they are complementary to each other.
“We have complete clarity on how they can be used independently and also in combination to provide users with knowledge and data,” he said.
Sage is Multilingual; Voice Capabilities Coming Soon
Sage is currently available in English but does a decent job when prompted in Hindi as well. The startup revealed that more languages could be added going forward.
Moreover, at Google I/O, the tech giant unveiled Project Astra, a first-of-its-kind initiative to develop universal AI agents capable of perceiving, reasoning, and conversing in real-time.
Project Astra is built on Gemini, and according to Kumar, the same voice capabilities would come to Sage as well.
“The adoption will rely heavily on voice interaction in natural language across different regions worldwide. You simply ask a question in your preferred language. This will be really beneficial for those less adept at typing or with limited literacy in agriculture,” Kumar adds.
However, he also emphasises that the startup will take action when the time is right, as it depends largely on their capacity to manage this transition effectively.
