Individual Presentation: Designing for Deliciousness: How AI Can Improve Sensory Outcomes for Sustainable Proteins

Sustainable protein manufacturers can implement all the innovations they want, but one fact will always be true: If it doesn’t taste good, your product won’t perform well on the market.The development of sustainable proteins, from precision-fermented dairy to novel plant-based or mycoprotein ingredients, demands an understanding of not only nutritional and environmental outcomes but also sensory performance. Taste, texture, and overall experience remain key drivers of consumer acceptance—yet sensory testing is often the longest and most resource-intensive part of R&D.In this talk, Sohum Patnaik, Machine Learning Engineer at Food System Innovations, will share how his team is using artificial intelligence to predict sensory and functional attributes of sustainable proteins directly from their formulation data. Drawing from his background in data science and machine learning, Sohum will explain how models trained on large ingredient and sensory datasets can forecast attributes such as creaminess, chewiness, and flavor intensity before a product ever hits the tasting lab as well as predicting market success.The session will explore the end-to-end pipeline for sensory prediction—from model development and validation to real-world applications across alternative protein categories. Case studies will demonstrate how these AI-driven models can be used to optimize texture and flavor, shorten prototyping cycles, and identify formulation levers that improve sensory appeal that can drive adoption.By integrating machine learning with traditional sensory and consumer science approaches, this research helps bridge the gap between computational prediction and human experience—supporting faster, more informed, and sustainable innovation in food design.