When a critical building system fails, every minute spent searching for information matters. For facility professionals responsible for keeping public buildings operating safely, even identifying the correct component can be a challenge, especially in aging or modified spaces where records may be incomplete or unavailable.

Researchers at Oklahoma State University’s College of Engineering, Architecture and Technology are working to change that.

Dr. Soojin Yoon and Dr. Ashish Kumar Pampana have developed EyeFM, an artificial intelligence–enabled facility management tool designed to help maintenance professionals identify building components using simple photographs and receive meaningful maintenance guidance in real time.

The idea behind EyeFM grew out of Pampana’s doctoral research, “New Paradigm of Facility Management Using Reverse Image Search.” His work explored how AI could help bridge a persistent gap in facility management: important knowledge scattered across manuals, work orders, online resources and personal experience.

Working alongside Yoon, assistant professor for the School of Fire, Construction and Emergency Management, Pampana envisioned a tool that could begin with something practical - a photo taken in the field - and transform it into actionable maintenance support.

“Many building components look similar, especially in mechanical spaces where equipment can be old, modified, unlabeled, or installed in complex configurations,” Pampana said. “In many cases, technicians may not have immediate access to detailed records while they are on site.”

EyeFM is designed for those realities.

Rather than functioning solely as an image-recognition system, EyeFM connects identification results to maintenance decision support organized into three research-informed categories: functionality, maintenance repair and replacement, and communication.

The functionality category helps users understand why a component matters, including how its failure could disrupt connected systems. MR&R provides technical references such as specifications, manuals and procurement-related context to support repair or replacement decisions. The communication category focuses on field-ready support, offering tutorials, videos and AI-driven troubleshooting guidance.

Development was further shaped by participation in the U.S. National Science Foundation Innovation Corps program, in which the research team conducted extensive customer discovery with industry professionals. Those conversations refined both the tool’s technical direction and its value proposition.

Customer feedback reinforced the need for a system capable of working with imperfect images and supporting quick, confident decision-making in the field. As a result, EyeFM evolved from a component identification concept into a broader maintenance intelligence platform.

“Users told us they needed more than recognition,” Pampana said. “They wanted help moving from ‘What is this?’ to ‘What does this mean, and what should I do next?’”

The underlying technology uses a reverse image search engine developed specifically for facility equipment, such as HVAC systems. The system leverages EfficientNet-B7, a pretrained deep learning model fine-tuned with a proprietary dataset of labeled facility component images, making it better suited to specialized building environments than general-purpose image recognition tools.

By shortening the path between observation and action, the research aims to support smarter, more resilient built environments, helping technicians respond effectively when problems arise and reducing the risk of larger system disruptions.

The research team envisions EyeFM evolving into a more comprehensive decision-support platform, expanding component coverage and integrating more deeply into facility workflows. Future directions include applications for training, knowledge transfer and digital facility operations across campuses and building portfolios.

More broadly, the project demonstrates how academic research can move beyond theory to create tools with real societal value.

That translation from research to practice was made much stronger through the support of OSU Facilities Management.

“Their willingness to engage with the project helped ground the work in real-world facility operations. We are especially thankful to Mr. Jeff Sweeden, director of OSU FM, for his continued encouragement and support, and to Mr. Greg Leeming, project manager, for his constant help throughout the project.”

By anchoring the research in real operational environments, the project’s focus naturally expanded beyond technical achievement. It became about impact, about ensuring the research could meaningfully support the people responsible for maintaining the buildings others rely on.

“What excites us most is turning research into something people can actually use,” Pampana said. “If we can help maintenance professionals do their jobs more safely, confidently and efficiently, that impact extends far beyond the technology itself.”