By Elizabeth Egan
Published October 22, 2026
Three faculty members in the School of Engineering and Applied Sciences have earned National Science Foundation CAREER Awards for projects ranging from advancing statistical methods and improving wireless networks to developing a new computational framework based on the concept of pore graphs.
The recipients include Xiaoyu Chen, assistant professor in the Department of Industrial and Systems Engineering; Kaihang Shi, assistant professor in the Department of Chemical and Biological Engineering; and Seyyedali Hosseinalipour (Ali Alipour), assistant professor in the Department of Electrical and Computer Engineering.
Xiaoyu Chen
In many modern scientific and engineering applications, data is naturally represented by probability distributions that describe variability within repeated measurements. However, many analytical methods and engineering techniques aggregate rich distributional information into summary statistics, discarding valuable information.
With a $500,000 grant, Chen will establish new statistical methods for analyzing data in which each observation is itself a probability distribution. Chen said that mathematically, such distributions form a curved space rather than a flat one since two distributions cannot be added or subtracted the way single numbers can.
A basic tool used in scientific discovery is linear regression. Chen said this can be as simple as a linear regression, which builds on the straight-line equation Y = aX + b.
However, Chen said when data contains repeated measurements, limited summary statistics or engineered features are no longer sufficient to capture the randomness and relationships in data. For example, if a person’s blood pressure is measured three times, each reading could be different. This creates multiple data points for the individual and raises the question of which reading should be recorded as the outcome. Averaging the three readings is a common answer, but the average hides how much the readings vary.
“I am trying to redefine the linear regression model by using the distributions as input and output so that when we are trying to make scientific discoveries, we are not making an inference of the model coefficient for the summary statistic but interpreting the significance of random distributions,” said Chen.
The project will establish the statistical foundations for a distribution-in-distribution-out analytical framework that can enable linear regression modeling and inference when both predictors and responses could be a single value or distribution-valued random variable.
Chen will explore how these statistical models can be applied to fields that are naturally represented by probability distributions, such as healthcare, biomedical engineering, manufacturing and even finance.
The project will also include an educational component. Chen will provide training opportunities to undergraduate, graduate and high school students in engineering statistics and biotechnology.
In addition, over the next five years, Chen plans to create an open-source software, pyDIDO, that implements the developed models, making them accessible for people to become familiar with the new models and implement them for use in their own discoveries.
Kaihang Shi
Metal-organic frameworks (MOFs) are a promising class of porous materials that have important applications across chemical separations, energy storage, semiconductor technologies and more. They contain intricate networks of nanosized pores that determine how molecules move through and interact with the material, ultimately controlling how the material performs.
According to Shi, computational modeling can help to increase understanding of these molecular behaviors, but current methods struggle to capture the complexity of the networks and the chemistry that takes place inside them.
With a $549,000 grant, Shi aims to develop a new computational framework based on the concept of “pore graphs,” a graph-theoretic representation of pore networks that will transform complex pore networks and chemistry into a unified mathematical object that can be quantified and analyzed.
By combining pore graphs with graph theory, molecular modeling and machine learning, Shi’s work will help improve the understanding of adsorption thermodynamics, molecular transport and confined material assembly in MOFs, all of which are difficult to address with conventional computational methods.
“Molecular thermodynamics, motion and dynamics are the foundation of how MOFs perform in practical applications,” said Shi, adding that determining how much hydrogen a material can store for clean energy applications and how selectively a material can capture pollutants from air are two examples of these practical applications. “Improving the understanding of how molecules move through complex pore networks can help to establish clearer links between MOF structure and material performance.”
Shi said this knowledge will advance the fundamental science of thermodynamics and transport in confined environments, while also leading to the design of better materials for carbon capture, hydrogen purification and critical mineral recovery technologies.
By the end of the project, Shi aims to establish a new graph-theory-integrated computational framework for studying the confined molecular phenomena in porous materials and to have advanced knowledge of different elements of MOFs.
Shi will also be incorporating education into his project through the development of a new course, “AI for Chemical Engineers,” at UB, collaborating with UB’s Center for Industrial Effectiveness to develop in-person summer workshops and an online course for engineering professionals, and developing a new AI and machine learning session for his department’s summer camp for high school students.
Ali Alipour
Alipour’s work sits at the intersection of wireless communications and artificial intelligence/machine learning (AI/ML) as he explores how wireless networks and intelligent algorithms can help each other.
Wireless networks can be used to train AI/ML models that in turn make wireless networks smarter and more efficient. This could lead to advancements such as better data rates, improved security, enhanced user quality of experience, and increased reliability in connections between devices and cellular infrastructure.
With a $550,000 grant, Alipour will advance AI/ML-driven tools for future wireless networks by establishing a framework for the Internet of Foundations Models (IoFM).
In the IoFM paradigm, Alipour said he envisions a future where foundation models, such as large language models, are deployed across everyday devices from smartphones to vehicles. Instead of relying on a centralized cloud-based intelligence, these models will be distributed through the network and operate close to where the data is generated.
“This eliminates the need to continuously transfer raw data to centralized servers, thereby significantly enhancing data privacy and reducing communication overhead,” said Alipour.
The research will focus on two main components. First, Alipour will work on developing models and theoretical frameworks to characterize how effectively foundation models can be adapted in distributed environments. Next, he will design communication- and computation-efficient fine-tuning algorithms, enabling large-scale models to be updated locally with minimal resource usage.
“Overall, the vision is to create a scalable, efficient and privacy-aware AI ecosystem, where foundation models operate seamlessly across distributed devices and continuously adapt to user needs,” said Alipour.
Another thing that Alipour said excited him about the project is its interdisciplinary nature. The project brings together wireless communications, AI/ML and optimization theory to develop a unified framework for the next generation of distributed AI systems.
Alipour also noted that he plans to integrate the outcomes of this research into his teaching, specifically incorporating modules on distributed fine-tuning of foundation models into his course “Machine Learning over Wireless Edge Networks.”
“This will expose students to cutting-edge topics at the intersection of wireless systems and modern AI, helping to train the next generation of researchers and engineers in this emerging area,” said Alipour.


