Practical guide for food scientists to build AI: data, algorithms, and applications

Abstract

Artificial intelligence (AI) is rapidly transforming scientific disciplines, yet its adoption in food science remains fragmented and often constrained to narrow application scenarios. This perspective provides a practical guide for food scientists to build effective AI models, grounded in the premise that sustainable progress depends on advancing three foundational pillars in tandem: (1) high-quality datasets to enable new application scenarios; (2) tailored algorithms addressing food science challenges; and (3) impactful applications that yield new scientific insights and/or industrial value. We discuss how each pillar contributes to AI in food science and the state-of-the-art methods (e.g., large language models, high-throughput platforms, physics-informed neural networks, and multimodal fusion) that support each pillar. Building on these insights, we propose a checklist to help food scientists plan, develop, evaluate, and deploy AI models. Aligning future work with these pillars and the checklist may foster more AI-driven discovery and accelerate innovation across the food sector.