UH Mānoa wins $2 million NSF grant for AI-driven food safety system

The University of Hawaiʻi at Mānoa has secured a $2 million grant from the National Science Foundation (NSF) to lead a four-year initiative developing AI-based technology for food safety. Starting September 1, 2026, the project aims to identify environmental threats to food production systems before they escalate into public health crises.

In partnership with the University of Nebraska–Lincoln, UH Mānoa is leading a collaborative effort supported by a total NSF award of $4 million. The technology integrates artificial intelligence with environmental sampling, chemical analysis via high-resolution mass spectrometry, and genetic analysis through metagenomics. This multi-layered approach monitors and identifies risks across agricultural systems, livestock, and aquaculture, with initial testing focusing specifically on aquaculture and beef cattle production systems.

The shift to proactive detection

The project addresses the limitations of current food safety protocols, which often identify contamination only after consumers have fallen ill. A primary motivator for the research is the challenge of detecting foodborne illnesses such as Cyclospora. In current public health investigations, Cyclospora is typically identified only after consumption, leaving a critical gap in preventative safety.

Tao Yan, the project's Principal Investigator and director of the Water Resources Research Center (WRRC), noted that current methods are reactive. "Right now we look at the Cyclospora problem, we know it after people have already consumed the food," Yan said. He explained that the AI-based tools being developed aim to detect these problems before they manifest into the larger outbreaks seen today.

Industry and consumer impact

By moving the point of detection from the consumer's plate to the production environment, the project seeks to fundamentally change how the industry handles food safety. Proactive detection reduces the risk of widespread illness and protects farmers from the devastating economic impact of large-scale recalls and contaminated harvests. The integration of AI allows for the processing of complex environmental data at a scale and speed that traditional sampling cannot match, potentially flagging anomalies in real-time.

Next steps for the initiative

Over the next four years, the research team will refine the integration of AI with genetic and chemical testing to ensure the system can accurately distinguish between benign environmental markers and genuine threats. While the initial focus remains on beef cattle and aquaculture, the long-term goal is to create a scalable framework applicable to a wider array of agricultural systems. The success of the project depends on the ability of these AI tools to provide reliable, early warnings that producers can act upon before food enters the supply chain.