Backed by a $2 million Bezos Earth Fund grant, Food System Innovations has launched an AI research lab to accelerate sustainable protein development through open-source tools.
Food System Innovations (FSI) has launched the Food Intelligence Lab, a new artificial intelligence (AI) research programme backed by a $2 million grant from the Bezos Earth Fund to accelerate sustainable protein product development. The initiative will develop open-source datasets, models and benchmarks to help food manufacturers shorten R&D timelines and improve taste, texture and affordability.
AI is already transforming fields like drug and materials discovery, but food still lacks the shared infrastructure needed to fully unlock the potential of AI in this space. We’re building tools to help food scientists iterate faster and create truly exceptional sustainable protein products.”
Anna Thomas, Director of Machine Learning at the Food Intelligence Lab and a computer scientist at Stanford University
The lab will combine sensory data from NECTAR with instrumental measurements, including texture profile analysis (TPA), pH and shear testing, to develop tools for formulation design and sensory prediction. FSI said it will make the datasets and models publicly available and work with companies, academic institutions and non-profit organisations to support product development.
“We believe AI can be a powerful accelerator for climate and nature solutions when it is paired with the right data, collaboration, and real-world applications, moving promising ideas into impact,” said Dr Amen Ra Mashariki, Director of AI at the Bezos Earth Fund. “We’re proud to support Food System Innovations’ work, a great example of the kind of opportunity the AI Grand Challenge sought to discover: bringing AI, open science, and sustainable protein research together to help unlock faster progress at scale.”
The programme has already demonstrated early results through a collaboration with Proxy Foods AI. Using an optimisation system called Expert-Guided Bayesian Optimisation (EGBO), the partners improved the sensory performance of a plant-based Greek-style yoghurt by 29 percent after 10 formulation iterations completed over five days. According to FSI, the final formulation matched an animal-based benchmark for consistency, creaminess and tanginess, while the AI system produced a stronger formulation than a professional food scientist working within the same time constraints.
Anna Thomas, Director of Machine Learning at the Food Intelligence Lab and a computer scientist at Stanford University, added: “AI is already transforming fields like drug and materials discovery, but food still lacks the shared infrastructure needed to fully unlock the potential of AI in this space.
“We’re building tools to help food scientists iterate faster and create truly exceptional sustainable protein products.”
Can AI solve the alternative protein taste challenge?
Alongside formulation optimisation, the lab is also developing tools to reduce the time and cost of sensory testing. Researchers have developed TasteBench, a publicly available benchmark and Kaggle competition featuring both food-level and molecular-level prediction tasks for evaluating AI models that predict how closely sustainable protein products resemble their animal-based counterparts. FSI said its best-performing model achieved sensory prediction accuracy comparable with the median NECTAR panellist.
The launch comes as alternative protein developers continue to face challenges improving taste and texture while managing lengthy and costly product development. According to NECTAR’s Taste of the Industry 2026 report, only 33 percent of participants rated dairy-free products as foods they “like” or “like very much”, with similarly low scores for meat alternatives.
Food systems generate around 26 percent of global greenhouse gas emissions, with livestock accounting for more than half of those emissions. FSI said improving the sensory quality and affordability of sustainable protein products could help drive wider consumer adoption and support food system decarbonisation.
“Food scientists shouldn’t have to spend months on trial-and-error to get texture, mouthfeel, flavour, and aftertaste right. At Proxy Foods AI, we’re building the technologies and operating infrastructure that help companies hit their R&D and sustainability targets faster, with the goal of eventually reaching 90 percent in silico and reserving the wet lab for the final 10 percent of validation and optimisation,” said Panos Kostopoulos, Founder and CEO of Proxy Foods AI.
“Every iteration, successful or not, adds value and compounds. Partnering with FSI’s Food Intelligence Lab to open-source these tools is how we accelerate those breakthroughs and ultimately change how we feed the planet for the better.”
Building the data foundation for AI-driven food innovation
Want to learn more about AI in food R&D? Join New Food’s upcoming free webinar, From ingredients to intelligence: building the data foundation for AI-driven food innovation, on 5 August at 3pm BST.
The webinar will explore how structured product data can support faster formulation, streamlined compliance and smarter decision-making, while providing a framework for evaluating AI opportunities before making significant investment decisions.
No comments yet