Scaling sustainable protein production: Integrating farmer knowledge with AI to optimize black soldier fly larvae cultivation

The transition toward a circular bio-economy requires scalable technologies for organic waste upcycling and sustainable protein production. Black Soldier Fly (BSF) farming has emerged as a promising approach for waste-to-resource conversion; however, its industrial expansion is constrained by biological variability and the lack of robust optimization tools suitable for decentralized production systems. This study presents a farmer-partnered Decision Support System (DSS) that applies machine learning to optimize larval biomass yield and nutritional quality, including protein and fat content, using real-world data collected across multiple farms. The proposed DSS follows a data-centric approach that integrates practitioner domain expertise with longitudinal experimental data while accounting for the biological variability inherent in organic side-streams. We evaluate cross-farm generalization compared to personalized modeling strategies and analyze how shared datasets can improve prediction reliability in heterogeneous farming environments. To support practical adoption, the system incorporates explainability and interactive visualizations that enable farmers to better understand how diet composition influences larval growth and nutrient outcomes. The results demonstrate that integrating data science with farmer knowledge can improve process consistency and resource efficiency in BSF production systems, contributing to reduced reliance on marine-derived feed inputs and supporting progress toward UN Sustainable Development Goals related to responsible production and environmental sustainability.