Carbon and data

Agriculture is the backbone of East Africa's economy, employing over 60% of the workforce and contributing 43% of Kenya's GDP. Yet the smallholder farmers who form this foundation, managing plots of 0.5 to 5 acres, are operating almost entirely blind. They have no affordable tools to monitor soil health, no way to track their carbon footprint, and no access to personalised guidance on how to improve their yields or farming practices. The few digital tools that exist are built for large-scale commercial farms in developed markets. They require expensive sensors, high-speed internet connectivity, or technical expertise that rural East African farmers simply do not have. As a result, over 97% of Kenya's 2 million smallholder farmers have never used a digital farm management tool. Climate change is making this worse. Unpredictable rainfall, rising temperatures, and shifting seasons are reducing crop yields across the region, yet farmers receive no real-time guidance on how to adapt. At the same time, growing global demand for carbon credits presents a significant economic opportunity for small farms, but farmers have no way to measure or prove their carbon sequestration impact. Shamba Agricultural OS directly addresses this gap. By combining a machine learning health assessment model, IPCC-standard carbon tracking, and Google Gemini AI recommendations into a single low-bandwidth web application, Shamba OS gives any farmer with a basic smartphone instant access.

Shamba Agricultural OS is a low-bandwidth web application that brings AI-powered farm intelligence to East African smallholder farmers for the first time. Built by Team GLAM from Mount Kenya University, Shamba OS combines machine learning, carbon science, and generative AI into a single tool that any farmer can use on a basic smartphone, with no sensors, no technical knowledge, and no account required. At the core of Shamba OS is a farm health assessment engine powered by a Random Forest machine learning model trained on 2,200 real crop data points from the Kaggle Crop Recommendation Dataset. The model achieves 99.8% accuracy in predicting farm health status across three categories: healthy, moderate stress, and critical. Rather than asking farmers to enter complex soil values they would not know, Shamba OS uses a simple region selector. When a farmer chooses their county from a dropdown, the system automatically fills in the appropriate soil and climate defaults for that region, derived from real East African agricultural data. The farmer simply selects their region, their crop type, and their farm size, and the AI does the rest. Once the assessment is complete, Shamba OS generates three personalised farming recommendations using Google Gemini 2.0 Flash AI. These recommendations are written in plain, farmer-friendly English with no technical jargon, and are tailored specifically to the farmer's crop type, health status, farm size, and carbon footprint. Every recommendation is practical and actionable, giving the farmer something they can do that same day. Carbon monitoring is built into every assessment. Shamba OS uses IPCC-standard emission factors to calculate each farm's carbon footprint from water usage, farm size, and energy consumption. The result is presented as a simple kg CO2 value alongside an A, B, or C grade, making it easy for any farmer to understand their environmental impact. Farms that achieve a Grade A score unlock eligibility information for micro carbon credit programmes, opening a new income stream that has previously been inaccessible to smallholder farmers. All assessment results are saved to the farmer's profile, enabling the historical trends feature. Farmers can track how their health score and carbon footprint have changed over time, see their best scores, and measure the impact of the recommendations they have followed. This transforms Shamba OS from a one-time tool into a long-term farm management companion. The full technology stack reflects a commitment to accessibility and performance. The frontend is built with React 19, TypeScript, Vite 6, and Tailwind CSS v4, optimised for low-bandwidth connections. The backend runs on Python FastAPI with JWT authentication and email OTP verification, ensuring farmer data is secure. The entire system is accessible from any browser, on any device, on a 2G connection, making it genuinely usable in rural East Africa where connectivity is limited. Shamba OS is not a prototype.