Yanet Niguse Tesfay
Speaker at AMLD Africa 2026
Yanet Niguse Tesfay is an AI/ML researcher, software engineer, and innovator passionate about harnessing artificial intelligence and machine learning to drive sustainable social impact. She is the creator of FarmBuddy, an AI-driven decision support tool that helps African smallholder farmers improve productivity through localized, data-driven insights. Yanet is also a Harvard WECode Fellow, Generation Google Scholar (EMEA), and Mastercard Foundation Scholar. Her work focuses on sustainable agriculture and food security, climate change, and empowering underserved communities through ethical, inclusive, and human-centered AI.
At AMLD Africa
WITS University, Johannesburg, South Africa
FarmBuddy
FarmBuddy is an AI-powered decision-support assistant designed to support sustainable farming among smallholder farmers in resource-limited settings. Farmers often struggle with poor crop selection, unpredictable weather, inefficient irrigation, and limited access to agronomic expertise, leading to low yields and high risk. FarmBuddy addresses these challenges by combining machine learning, fuzzy logic, and conversational AI into a single, accessible platform. The system integrates three components. First, a machine learning–based crop recommendation model trained on soil and climate features (NPK, pH, temperature, rainfall, and humidity) achieves 99.55% test accuracy using a tuned XGBoost classifier. Tree-based models were selected for their ability to capture non-linear relationships and provide interpretable feature importance, highlighting potassium, rainfall, and humidity as key drivers. Second, a fuzzy logic–based smart irrigation module generates context-aware sprinkling recommendations from soil moisture, temperature, and humidity inputs, supporting water efficiency and climate-smart agriculture. Third, a 24/7 chatbot advisory system delivers localized guidance on crop selection, pest control, and fertilizer use, trained on KALRO agricultural extension materials. FarmBuddy is deployed through a user-friendly Streamlit web interface, with Joblib for model persistence and PyQt5 for irrigation visualization. This talk presents the system design, evaluation results, and deployment insights, alongside reflections on interpretability and practical adoption. Future work includes integrating SHAP/LIME for explainability, incorporating real-time sensor data, and validating recommendations with agronomists. FarmBuddy demonstrates how applied machine learning can deliver practical, trustworthy tools for sustainable agriculture across Africa.
