Blending Materials Science, Nutrition Data, and AI

Not too far in the future you may find yourself asking the personal nutrition assistant on your phone to help you decide what to have for dinner. Based on your current protein, iron, cholesterol and sodium levels among other things, your assistant may help you choose between a hearty salad or a steak—perhaps one that remains perfectly fresh after a week at room temperature, because it’s been preserved with flash heating. If instead, your assistant nudges you toward a salad and recommends tossing in a few berries to up your antioxidant levels, those berries may have been washed in a solution derived from shrimp and crab shells to remove the pesticides and keep them fresh. Whichever meal you choose, when you’ve finished, you may wrap the leftovers in a biodegradable, sustainably made “plastic” wrap.

If any one of those things happens, Peihua Ma will have had a hand in it. As an AGNR post-doctoral associate, and a food and nutrition science Ph.D. alum, Ma has been collaborating with materials science engineers and applying the tools of artificial intelligence to a wide range of innovations in the food and nutrition arenas.

When faced with an intriguing food science question—like what compounds can be used to clean chemical residues from bumpy fruits and vegetables, or how to formulate a better type of plastic wrap—Ma collects as much data as he can about various materials and asks a computer to figure out the answer. He had a robot make 3,000 different AI-generated formulas for plastic wrap before settling on a few successful candidates.

Ma’s focus is on using deep learning and computer algorithms in food science, but those bastions of AI weren’t part of his initial research when he arrived at UMD more than six years ago. His primary objective then had been to earn his Ph.D. in Food Science with Professor Qin Wang, who is developing nano-sensors for food safety and studying nanomaterials for delivering medicines and bio-active compounds.

But when COVID shut down most of Ma’s laboratory work, he had to rethink his purpose here.

“I was being paid as a research associate,” Ma recalled, “And because campus was closed, I had to find something to work on so I could get paid.” He reached out to materials science colleagues in the Clark School of Engineering and began asking questions about how their work in artificial intelligence could help answer food science questions.

It was the perfect time to spend long hours at home in front of a computer, a place where he was very comfortable. From his kindergarten years all the way through high school, Ma spent much of his free time, and by his own admission too much of his study time, playing computer games, where his ambitions led him into coding. His singular focus and dedication suggested he was on track to study computer sciences. But his parents worried a path that encouraged him to spend his life in front of a computer would be unhealthy, so they steered him toward his other passion: food. They hoped he would learn about nutrition and possibly curb his appetite for barbeque and fried food.

But Ma never really let go of either of his passions. While pursuing his undergraduate degree in food science in China, Ma and his friends opened a barbeque restaurant, and by the time he came to the U.S. for graduate school, he was a skilled host who had become known for throwing multiple Chinese barbeques a year, often feeding 50 professors, students, and post-docs at a time.

He has also become somewhat of a connoisseur of world foods, having dined in over 100 Michelin star restaurants and frequented a bevy of street vendors in nearly 30 countries and 47 U.S. states. It's been an odyssey he refers to as “eating my way around the world.” He said his wealth of experience dining out is, in part, what inspired him to work on developing an AI food and nutrition assistant.

With an internship at the U.S. Department of Agriculture, Ma began building an AI model with Open AI, the company behind ChatGPT, to answer consumer questions about food and nutrition with science-based, authoritative information. The model is still in development, but Ma hopes to take the project even further on his own, creating a personalized nutrition AI.

“I am trying to see if we can use a large language model—an LLM like ChatGPT—to learn about individuals and synthesize that information with analysis of food composition and make personal recommendations,” Ma explained. “Because right now, food labels are very general, and what is good for one person’s body, maybe someone who has high blood sugar or heart disease, may not be good for somebody else.”

Ma also wants to develop medicinal compounds from multiple ingredients using AI. It is well understood that some compounds work better in collaboration, like calcium which absorbs in the body better when taken with vitamin D. But Ma, inspired by the complex soups his mother makes to cure different ailments, believes that more powerful, undiscovered medicines may be discovered by combining ten or more ingredients. He plans to use AI to develop recipes and robotic cooks to create and test them for medicinal properties in laboratory cell cultures.

So, if, in the not-too-distant future, your AI nutrition advisor recommends the oxtail soup to improve your liver function, or the pharmacist offers you a special compound to reduce your risk of cancer, there’s the possibility that Peihua Ma may have had a hand in it.

by Kimbra Cutlip : Momentum Magazine Winter 2025