CSIRO joins global effort to turn agrifood waste into competitive protein using AI and fermentation

Australia’s national science agency, CSIRO, partnered with the University of Leeds in the UK on a global research project aimed at repurposing agrifood waste into high-quality protein, as concerns around food loss, sustainability, and food security continued to intensify.

• CSIRO and the University of Leeds partnered on a US$3 million, two-year project to develop an AI-powered fermentation tool for converting agrifood waste into protein.

• The project focused on vegetable waste, grain byproducts, and dairy processing sidestreams to produce microbial protein for food and animal feed.

• The work was supported by the Bezos Earth Fund as part of its AI for Climate and Nature Grand Challenge.

Announced on 14 January 2026, the project brought together expertise in fermentation science and artificial intelligence to address one of the food system’s most persistent inefficiencies: the loss of nutrient-rich materials that could otherwise be used to produce food. In Australia alone, more than seven million tonnes of food were wasted each year, representing around one third of all food produced nationally.

Globally, the challenge was even larger. Billions of tons of agrifood byproducts and surplus crops were lost annually due to supply chain inefficiencies, quality standards, or processing limitations, despite containing valuable nutrients. The CSIRO-led project aimed to demonstrate how those materials could be redirected into protein production at scale.

The research centered on using fermentation to transform agrifood waste into microbial protein powder suitable for human food or animal feed. By combining fermentation with an AI-powered decision tool, the team aimed to calculate optimal processing conditions across different waste streams, reducing trial-and-error experimentation and improving the economic viability of upcycled protein.

The project examined three major categories of agrifood waste. These included vegetable crops that had been damaged or left unharvested, grain byproducts such as canola meal and brewer’s spent grain, and byproducts generated during cheese production. Each stream presented distinct technical and logistical challenges, particularly when it came to consistency, contamination risk, and nutrient variability.

CSIRO Project Lead Dr Kai Knoerzer said the scale of global food loss made the issue impossible to ignore. “Globally, billions of tonnes of nutrient-rich material are currently being lost each year,” he said. “Working with our colleagues internationally, this project will combine AI, fermentation science and real case studies to support industry to turn that waste into sustainable protein at scale.”

While fermentation had been used for thousands of years to preserve foods such as bread, cheese, and wine, the researchers argued that its role in the modern food system could extend much further. According to CSIRO, advances in microbial fermentation and process control meant that waste streams could now be converted into consistent, functional protein ingredients rather than low-value outputs.

Dr Knoerzer said fermentation had the potential to play a much larger role in strengthening food security in the 21st century. “When fermented using yeast, for example, food waste can be transformed into valuable products within the framework of a circular bioeconomy,” he said.

A key focus of the project was economic competitiveness. The researchers emphasized that upcycled protein would only meaningfully contribute to global food security if it could compete with conventional protein sources on price, not just sustainability credentials.

Dr Nicholas Watson, Professor of Artificial Intelligence in Food at the University of Leeds, said affordability was central to the project’s goals. “To truly impact global food security, upcycled protein couldn’t just be a niche alternative – it had to compete on price with what is already on the supermarket shelf,” he said. “We are excited to work with CSIRO and partners across the globe to bridge that gap, launching an AI platform to support the fermentation of agri-food waste.”

The AI tool under development was designed to integrate real-world data from fermentation trials with predictive modeling. By calculating optimal conditions for different feedstocks, the platform aimed to help producers determine whether a given waste stream could be converted into protein efficiently enough to justify commercial deployment.

Support for the project came from the Bezos Earth Fund, which backed the work through its AI for Climate and Nature Grand Challenge. The US$3 million initiative formed part of a broader US$100 million program focused on using artificial intelligence to address climate change and biodiversity loss.

Dr Amen Ra Mashariki, Director of AI and Data Strategies at the Bezos Earth Fund, said the project illustrated how AI could be applied responsibly to environmental challenges. He said the work showed how artificial intelligence, when guided by science and local knowledge, could strengthen environmental action and ensure its overall impact on the planet was positive.

Running over two years, the project aimed to generate practical tools and case studies that industry could use to assess the feasibility of upcycled protein production under real-world conditions. Rather than focusing solely on laboratory success, the team emphasized the importance of scalability, cost control, and integration into existing food and feed supply chains.

By combining AI, fermentation, and diverse agrifood waste streams, the CSIRO and University of Leeds collaboration reflected a growing shift toward circular approaches to protein production. The researchers said the ultimate objective was not simply to reduce waste, but to turn it into a reliable source of nutrition that could contribute meaningfully to global food security.