As a UF student and NVIDIA intern, I’m using agricultural AI to help transform global food production
When people hear that I am completing my doctorate in agricultural and biological engineering at UF, they often assume my work is confined to farmland, crop yields and tractors. But applying advanced AI to agriculture requires solving some of the most complex computing challenges in existence — challenges that directly impact how technology interacts with the real world.
In everyday life, computers usually see the world as flat, two-dimensional photos, but our physical environment is three-dimensional, unpredictable and constantly changing. The goal of my research is to bridge that gap — to teach AI not just to look at an image, but to truly understand depth, spatial layout and physical objects in three dimensions.
Under Dr. Won Suk Lee’s guidance at UF, I learned to approach research from first principles. Rather than focusing only on a particular crop, sensor or application, I was encouraged to identify the fundamental technical challenge, understand the assumptions behind different methods and develop solutions that could generalize beyond one specific problem.
Over the past five years at UF, this approach enabled me to publish six first-author journal papers across top publications like Computers and Electronics in Agriculture, Smart Agricultural Technology and Precision Agriculture.
The work I am most proud of is “PheMuT: A Phenology-Informed, Multi-Modal Time-Series Model for Strawberry Yield Forecasting.” The objective of this research is to forecast strawberry yield by combining knowledge about the fruit’s life cycle with numeric, textual and visual information. It addresses a problem with direct practical value, helping growers make better-informed decisions about harvesting, labor, logistics and resource allocation.
The collaborative research environment within the UF Institute of Food and Agricultural Sciences has played a foundational role in shaping my engineering mindset. Participating in journal clubs, such as the one organized by Dr. Charlie Li’s laboratory, and collaborating with researchers like Dr. Yiannis Ampatzidis and Dr. Henry Medeiros taught me that complex engineering problems are rarely solved in isolation.
Receiving the 2025 UF ABE Graduate Student Mentoring Award allowed me to guide undergraduate researchers, which significantly sharpened my technical communication, project management and leadership skills — abilities that are just as vital in an industry research environment as coding proficiency.
Transitioning from an agricultural engineering lab at UF to working as an intern at an industry AI powerhouse like NVIDIA was a natural evolution because outdoor agricultural environments are inherently chaotic. Agricultural computer vision is widely recognized as a difficult real-world challenge.
Outdoor environments contain constantly changing illumination, weather conditions, shadows, background clutter, occlusion, seasonal variation and highly complex biological structures.
Working in these environments taught me that strong performance on a clean or carefully curated dataset is not enough; a model must remain reliable when the scene is incomplete, noisy, ambiguous or different from the training data.
With the Metropolis team at NVIDIA, I help AI models better understand the physical world and spatial information. At a high level, my work focuses on using vision-language models to detect and reason about 3D defects in industrial environments. Whether I’m helping a farm manager predict food supply or helping an engineer spot a hidden structural flaw on a printed circuit board, giving AI spatial intelligence makes our physical systems safer, more efficient and more reliable.
A typical day for me involves reviewing recent research, designing experiments, developing and evaluating model pipelines, analyzing failure cases and discussing results with the team. The broader goal is to create an experience similar to interacting with ChatGPT: a user could provide several images of an object or industrial component, ask questions in natural language, and receive an answer that identifies, localizes and reasons about a defect in 3D space.
As I prepare to complete my Ph.D. in December 2026, this internship at NVIDIA has been a defining milestone. It has expanded my understanding of what is possible in AI research and confirmed my long-term goal: to continue working at the intersection of computer vision, multimodal AI, spatial intelligence and robotics to build systems that can seamlessly understand and interact with the physical world.
My advice to fellow UF students is to recognize that a specialized research domain is not a limitation. Look beyond the immediate application, find the fundamental challenge underneath it, stay curious and have the courage to apply your skills to the world's biggest stages.
Jing (Zijing) Huang is a Ph.D. candidate in agricultural and biological engineering at the University of Florida Institute of Food and Agricultural Sciences. Currently interning with the Metropolis team at NVIDIA and planning to join NVIDIA full-time post-graduation, his award-winning research focuses on leveraging computer vision, robotics and AI to solve complex spatial challenges in both agriculture and industry.