In Harvard University’s dining halls, a small camera watches as food is scraped from trays into the trash. The setup looks unremarkable, but it’s part of a high-tech effort at waste reduction — one that treats every discarded spoonful as data.
The system, called Winnow Vision, uses artificial intelligence to recognize what food is being thrown away, weigh it, and calculate its cost. The goal is to help chefs cook more accurately and waste less.
Food waste is a sprawling global issue. The United Nations estimates that one-third of all food produced is lost or wasted each year, even as hundreds of millions go hungry. In the United States alone, the think tank ReFED estimates that food waste costs businesses more than $100 billion annually.
For a university that serves roughly 6,600 undergraduates three meals a day across 13 dining halls, diminishing what ends up in the trash could amount to major savings.
“Winnow Vision is the realization of what we set out to do from the beginning,” said Marc Zornes, the company’s co-founder. “It’s an intelligent camera that sits on top of where food is thrown away. Our technology is trained to identify what that food is, putting a cost to that waste, and automatically giving that insight to the chef.”
This fall, Harvard University Dining Services began testing the system in three dining halls — Adams, Currier and Annenberg — to track exactly what gets wasted in the kitchen and in the dining hall. The cameras photograph food that’s discarded, while a scale measures the weight. Staff enter the day’s menu so the software can link scraps to specific dishes.
It is a feedback system for cooks. When too much of one dish ends up in the bin, the data show it. When students devour another, the kitchen knows that, too.
The early results of the pilot program have been enlightening. “The number one actually is rice,” said Martin Breslin, Harvard’s director for culinary operations. “We saw it across the three locations and that was a very easy fix — like immediate.”
Some of the technology’s growing pains are more subtle. “Did the vision AI from Winnow truly capture the ingredient or the item, or is it something else — like, is it macaroni and cheese, or is it chicken noodle soup?” asked Smitha Haneef, managing director of dining services. “There have been some errors.” Each night, staff review the system’s reports and correct any mistakes, which in turn trains the software to improve.
Even with occasional misfires, the implementation of the system represents a larger shift in how Harvard thinks about food waste. The university already composts 26,500 pounds of food each week — more than half a million pounds a year — and donates untouched leftovers to local shelters through a nonprofit program called Food for Free. But composting, while environmentally responsible, doesn’t solve the core issue of overproduction.
Harvard is looking upstream, toward prevention. The dining service has even shortened its menu cycle from four weeks to three, reducing the number of ingredients kept in storage and focusing on dishes that students are more likely to eat. “The four-week cycle really stretches and adds lots of items to inventory and so on, which can actually add to increasing waste,” Breslin said. “So the three-week cycle menu is actually more efficient.”
The Winnow pilot aims for a 15% reduction in food waste at the kitchen production level by semester’s end. If successful, the system could expand across all dining halls next year.
Harvard’s trial is part of a wider institutional effort to meet its Sustainable and Healthful Food Standards, which emphasize reducing environmental impact and improving food literacy. It’s also part of a global movement: more than 3,000 commercial kitchens worldwide now use Winnow’s technology, from hotels in Barcelona and Cancun to corporate cafeterias in London and New York. Many report cutting food waste by up to half and saving hundreds of thousands of dollars in the process.
The Hotel Barcelona Princess installed the technology in late 2024 as part of its corporate sustainability program, according to Winnow, and within eight months, the hotel cut its food waste by 63%, far exceeding its original 40%goal. Kitchen managers said the AI data made it clear which menu items routinely went untouched and allowed them to trim production without affecting guests’ choices.
In Cancun, the Royalton CHIC resort saw similar results. At its main buffet restaurant, food waste dropped 56% within six months of using Winnow. The savings were tangible: less money spent on overproduction, less waste hauled away, and a smaller carbon footprint to report to investors. Management credited the system’s daily readouts — a visual ledger of what ended up in the trash — for changing long-standing kitchen habits almost overnight.
At the Renaissance Newport Beach Hotel in California, Executive Chef Paul Bauer reported a 31% reduction in food waste just five months after installing the system. Across all locations, the pattern is consistent: once cooks can see the waste in numbers and images, they start adjusting automatically. It’s not the technology alone that saves food, managers say — it’s the awareness it forces every day at the garbage receptacle.
While the AI may not always know the difference between pasta and soup, it’s already changing how kitchens sees its work.
