Nathan Kahl

lighter-than-air robotics

Autonomous blimps take flight at EagleBank Arena: 2nd annual Airborne Robotics Cup inspires future STEM innovators

Nathan Kahl — Tue, 08 Sep 2026 19:04:00 +0000

Autonomous blimps take flight at EagleBank Arena: 2nd annual Airborne Robotics Cup inspires future STEM innovators Nathan Kahl Tue, 09/08/2026 - 15:04

In This Story

The floor of EagleBank Arena buzzed with high-tech aerial excitement on August 15 as 91�� hosted the culmination of the second annual Airborne Robotics Cup (ARC 2026). The STEM outreach competition ran an intense 12-hour tournament featuring dynamic, lighter-than-air autonomous blimp “soccer” matches.

Lighter-than-air vehicles prepare for battle. Photo credit Yidi Huang.

ARC was founded and organized by Ningshi Yao, assistant professor in the Department of Electrical and Computer Engineering (ECE) and director of the Control, Intelligent Autonomy, and Optimization (CIAO) Lab. It was co-organized by the Capital Youth Outreach Club (CYOC).

The initiative serves as a core component of Yao’s National Science Foundation (NSF) CAREER Award. As part of the award’s integrated education plan, it translates university research in robotics, control theory, and autonomous systems into transformative hands-on learning for K-12 students, with vital venue, equipment, and organizational backing from the College of Engineering and Computing (CEC) and the Mason Autonomy and Robotics Center (MARC).

The tournament received enthusiastic support and on-site participation from CEC leadership and its academic departments. Ariela Sofer, divisional dean of the Volgenau School of Engineering, kicked off the morning session by encouraging students to pursue curiosity and explore the expansive possibilities within engineering. During the evening awards ceremony, Amarda Shehu, associate dean for research, celebrated the students’ accomplishments and underscored the transformative impact of early engagement with robotics and artificial intelligence, alongside Alex Sprintson, ECE chair, who recognized the exceptional ingenuity and technical rigor demonstrated by the competitors.

Alex Sprintson, ECE department Chair; Ariela Sofer, divisional dean of the Volgenau School of Engineering; student Ningshi Yao; student Aaron Liu; and Capital Youth Outreach Club co-organizer Jun Zhao. Photo credit John Liu.

Adding to the strong departmental support, Pelin Kurtay, ECE associate chair, attended to engage with participants and cheer on the young roboticists. Leigh McCue, chair of the Department of Mechanical Engineering, joined the morning awards session to deliver encouraging remarks on multidisciplinary engineering and present awards to the winning teams across the morning’s two competition groups.

Building on last year's debut success, ARC expanded exponentially in 2026. The program received more than 180 applications, and through a highly competitive selection process, 16 talented finalist teams comprising more than 80 middle and high school students earned spots in the tournament.

Rather than a standalone one-day competition, ARC represents a comprehensive two-month summer engineering program. Over the past seven weeks, students dedicated their weekends to four intensive in-person training sessions and four hands-on practice sessions at MARC. Led by Yao, alongside ECE PhD student Ahnaf Asif Khan, graduate student Yojan Gautam, and former George Mason Aspiring Scientists Summer Internship Program participants Adam Bhatti and Harry Yu, the mentorship team taught students foundational mechatronics skills from the ground up, including soldering, assembling basic blimp kits, embedded programming, and custom 3D printing.

Student leadership and authentic engineering problem-solving took center stage throughout ARC2026. Reflecting on the engineering hurdles of lighter-than-air robotics, Lucas Liu, a junior from Thomas Jefferson High School for Science and Technology and captain of Team 14, noted, "Working with blimps taught our team very quickly that every single gram matters. Through constant iteration, we redesigned parts and switched to lighter microcontrollers, bringing the total weight down to 70% of the original."

Aaron Liu, captain of Team 12 and representing his team from Luther Jackson Middle School, echoed the value of hands-on troubleshooting, "Building and flying a robot is a lot more complicated than watching one fly. Engineering problems don’t always happen when it's convenient, and solutions don't come linearly. You just have to keep going one step at a time," said Aaron.

Reflecting on the broader impact and purpose of the initiative, Yao emphasized the long-term growth of the participants. "ARC is far more than just a competition," said Yao. "It is a launchpad for lifelong skills, including problem solving, critical thinking, teamwork, leadership, and resilience. Today is a celebration of taking ideas from concept to reality, embracing setbacks, and learning through hands-on trial and error."

Winning teams celebrate their achievements alongside faculty and mentors during the awards presentation. Photo credit Yidi Huang.

That transformative experience resonated strongly with families across the community. "Your continuous support, patience, guidance, and extra effort meant so much to them and to us as parents," shared Sisi Jou, parent of Team 15 competitor Skylon Jou. "Skylon already said he definitely wants to participate again next year and hopes to come back even stronger!"

Building on the tremendous energy and success of this second edition, the CIAO Lab and its partners are already preparing to expand ARC for summer 2027 by increasing team capacity, broadening community outreach, and continuing to cement George Mason as a premier hub for youth robotics innovation.

Topics

Electrical and Computer Engineering

STEM outreach

EagleBank Arena

robotics

Mason Autonomy and Robotics Center

National Science Foundation

Students

Designing deeper data center defense

Nathan Kahl — Fri, 04 Sep 2026 01:25:30 +0000

Designing deeper data center defense Nathan Kahl Thu, 09/03/2026 - 21:25

In This Story

Kai Zeng

Liling Huang

As artificial intelligence (AI) drives rapid growth in data centers, the facilities powering the technology are becoming more complex, more valuable, and potentially more vulnerable to cyberattacks. Two 91�� researchers want to help close security gaps before attackers can exploit them.

Liling Huang. Photo by Ron Aira/Office of University Branding

Kai Zeng and Liling Huang, a professor and associate professor, respectively, in the Department of Electrical and Computer Engineering, are developing new approaches to protect AI data centers from threats ranging from insiders stealing valuable AI models to cyberattacks against the systems controlling a facility’s power and cooling.

Zeng and Huang received $100,000 from the Virginia Innovation Partnership Authority through the Commonwealth Cyber Initiative Northern Virginia Node for “Defense-in-Depth Cyber-Physical Security for AI Data Centers.” The one-year project examines vulnerabilities that traditional cybersecurity approaches may not fully address, including the increasingly complicated intersection of information technology, operational technology, and building-management systems.

“It’s a very urgent and emerging topic,” Zeng said. “Just look at how many new AI data centers are built every year, every month.” The issue has particular resonance in Virginia, which has the world’s largest concentration of data center infrastructure and is rapidly becoming a hub for large-scale AI computing.

AI data centers present new challenges because the systems differ from conventional data centers, Huang said. Operators in traditional facilities have years of experience understanding normal activity, making anomalies easy to identify. With AI infrastructure, that baseline is still developing.

“With AI data centers, we are still learning about their operational behavior and how it affects both behind-the-meter electrical systems and the broader power grid,” Huang said. “This uncertainty makes cybersecurity more challenging because, when abnormal behavior occurs, it can be difficult to distinguish a cyberattack from a legitimate equipment or power-system issue.”

The researchers will investigate three areas. One is communication among the clusters of graphics processing units (GPUs) used for AI computing. Their extremely fast connections can make monitoring activity difficult. “It’s harder to monitor the traffic exchange there between the GPUs, which could be a kind of blind spot from a security perspective,” Zeng said.

Another focus is protecting what’s known as “checkpoints,” snapshots of an AI model’s state created during training. Training a large AI model can take weeks. At intervals, the system saves the model's current state, including what it has learned so far. That saved state is the checkpoint, which can contain valuable intellectual property, making them potential targets for tampering or insider attacks. The researchers plan to explore whether data collected from systems throughout the data center, known as telemetry, can help verify that a checkpoint is authentic.

Kai Zeng. Photo by Office of University Branding

The third area involves the boundary between IT systems and operational technology, including systems controlling power and cooling. “Someone does not necessarily have to directly hack into the server, because they can just disrupt the building management system,” Zeng said.

The project benefits from the researchers’ complementary expertise. Huang explained that cybersecurity practices have traditionally been more mature in information technology systems, while operational technology has emphasized reliability, safety, availability, and engineering performance. Zeng brings cybersecurity expertise, while Huang contributes expertise in AI data centers, electrical and mechanical systems, and the power grid. Together, they will work to strengthen cybersecurity protections for the operational technology systems of AI data centers.

The team plans to produce a data center threat model, prototype security technologies, experimental results, and at least one peer-reviewed publication. The seed project is also intended to position the researchers for larger federal and industry-funded projects.

Zeng hopes the initial investment can provide the foundation for something considerably larger. “We can apply later to turn this $100,000 project into, potentially, $1 million,” he said. “Start small, but scale up, and scale big.”

Topics

Cyber Security

Commonwealth Cyber Initiative (CCI)

Electrical and Computer Engineering

data centers

George Mason researcher brings critical AI literacy program to Prince George’s County

Nathan Kahl — Mon, 31 Aug 2026 22:18:35 +0000

George Mason researcher brings critical AI literacy program to Prince George’s County Nathan Kahl Mon, 08/31/2026 - 18:18

As artificial intelligence (AI) rapidly transforms society, 91�� researcher Thema Monroe-White is helping prepare the next generation not just to use the technology, but to navigate it thoughtfully and responsibly. Her work empowers students to understand AI's opportunities, limitations, and consequences, giving them the agency to make informed choices in an increasingly AI-driven world.

Thema Monroe-White. Photo provided

Monroe-White, an associate professor with appointments in the Department of Computer Science and the Schar School of Policy and Government, is part of a new $1 million initiative aimed at helping Prince George’s County students understand not only how to use AI, but how AI works, who it affects, and when it should and should not be used.

The project is funded by the Pull Up Fund, a philanthropic organization focused on supporting youth in Prince George’s County through its Youth AI Mastery and Proficiency initiative. George Mason is working with Ashley Griffin Gilchrist at Bowie State University and Evan Shieh at the Young Data Scientists League (YDSL) on the project. The initial award covers a one-year pilot, with the potential for two additional years of funding.

For Monroe-White, “AI mastery” shouldn't simply mean becoming proficient at using the latest AI platform. “To have or gain AI mastery requires not that you know how to prompt any given model,” she said. Instead, students should understand some of the technical machinery underneath AI, including how models generate outputs, as well as the human and social systems surrounding the technology.

That approach, known as critical AI literacy, asks students to examine issues such as algorithmic bias, labor, privacy, surveillance, environmental impacts, and the human work involved in creating AI systems. It also encourages them to question the technology itself rather than treating AI as a mysterious digital oracle.

“What does it mean? What is it doing? How is it operating? How is it affecting my life?” Monroe-White said.

The Prince George’s County initiative builds on work Monroe-White recently completed with New York City Public Schools and YDSL. Through the Exploring Equity in AI fellowship, educators explored ways to bring critical AI literacy into K-12 classrooms. The program defines AI literacy as encompassing not merely using AI, but “critiquing, researching, evaluating, (re)designing, and reimagining AI.”

The New York effort included topics ranging from AI and consent and algorithmic auditing to data centers and questions about how “intelligent” artificial intelligence actually is. Monroe-White and YDSL Executive Director Evan Shieh served as facilitators of the Critical AI component. Monroe-White is also an external research collaborator with YDSL.

After completing the New York pilot and a public showcase in June, Monroe-White saw the Prince George’s County project as an opportunity to expand the approach closer to home.

“We wanted to expand it and keep the work going,” she said. “And so that’s what we’re doing with Maryland and this county.”

Ultimately, Monroe-White wants young people to understand the technology well enough to make informed decisions about it themselves. Students, she said, should graduate “armed and equipped with the knowledge that most don’t have that demystifies AI,” allowing them to engage in meaningful conversations about when AI is appropriate and what consequences its use might have.

“It’s more of an agency-based training,” Monroe-White said. “It’s lifting up their own individual agency so that they’re empowered in this revolutionary moment.”

In This Story

Thema Monroe-White

Explore AI at George Mason

Topics

K-12 Partnerships

Schar School of Policy and Government

GCI-Education

Young engineers compete, building from the ground (and) up

Nathan Kahl — Wed, 19 Aug 2026 17:41:52 +0000

Young engineers compete, building from the ground (and) up Nathan Kahl Wed, 08/19/2026 - 13:41

Over the course of two Saturdays in August, 91�� electrical and computer engineering Assistant Professor Ningshi Yao tested the idea that competition yields strong results. She designed two very different engineering challenges for local pre-college students, one sending student-built ground vehicles racing across a soccer field and the other taking to the skies.

"We want to develop a course or training program that's more hands-on and learning through hands-on projects," Yao said.

“Whether it’s on the ground or in the air, engineering isn't just about formulas on paper—it's about learning through iteration, embracing setbacks, and building things that work in the real world," Yao said. "These programs give young minds early exposure to authentic engineering challenges, empowering them to discover their passions, build resilience, and see themselves as the next generation of roboticists and innovators.”

High school teams test autonomous blimp designs during hands-on practice sessions at MARC. Photo provided.

The first event, the NEXT Competition on August 7, capped a seven-week summer experience for 16 high school students participating in George Mason's Aspiring Scientists Summer Internship Program (ASSIP). The students helped Yao pilot a new mechatronics curriculum developed through a three-year, $750,000 grant from the Office of Naval Research (ONR).

The curriculum uses a flipped-classroom model. Students learn concepts outside the lab through lectures and select online videos, then spend their time together applying those concepts. Lessons cover topics ranging from sensors and mechanical drive trains to projectile motion and radio communication. "The class time will be when they come into the lab and really build a ground vehicle," Yao said.

The culminating competition tests far more than whether the vehicles move. Teams face off for 90 minutes on a soccer field scattered with 200 tennis balls. Their robots must collect the balls and shoot them into designated buckets while also contending with opposing vehicles. Teams can even fire balls at competitors to temporarily knock them out of action. That forces students to consider offense, defense, communications, vehicle maintenance and strategy while responding to problems they couldn't predict beforehand.

"They need to quickly fix things," Yao said. "They also need to think about communication, because the range is easily more than 150 meters. Wi-Fi and Bluetooth may not be enough, so they need to learn radio communication."

The ASSIP students were more than participants. Because the ONR grant was awarded only in April, they served as the first test group for a curriculum ultimately intended to stand on its own.

"We're using the high school interns as a pilot to test this curriculum," Yao said. "Meanwhile, they can identify the research problems they want to work on."

Just one week later, Yao shifted from ground robots to flying ones. On Saturday, August 15, more than 100 students were at EagleBank Arena for the second annual Airborne Robotics Cup (ARCup), a competition Yao founded to engage middle and high school students in robotics and artificial intelligence.

Teams of students in grades 8 through 12 spent approximately two months designing, building, and programming lightweight robotic blimps before competing in a soccer-style game. The blimp platform introduces students to CAD, 3D printing, programming, aerodynamics, stability, buoyancy, and robotics while remaining safe enough to operate around competitors and spectators.

Reminiscent of the collegiate-level Defend the Republic competition, students use the lighter-than-air devices to play a game reminiscent of Harry Potter’s Quidditch. Students receive mentorship from members of Yao's robotics research team, and top performers may pursue research internships.

Yao's commitment to ARCup recently received another boost. Her new NSF CAREER Award includes support for continuing and expanding the competition as part of the educational component of her research.

In This Story

Ningshi Yao

Learn more about ASSIP

Topics

competitions

Department of Electrical and Computer Engineering

EagleBank Arena

Defense and Security

Students

George Mason prepares to launch new PhD in Cybersecurity

Nathan Kahl — Mon, 17 Aug 2026 16:57:39 +0000

George Mason prepares to launch new PhD in Cybersecurity Nathan Kahl Mon, 08/17/2026 - 12:57

In This Story

Paulo Costa

Tanvir Arafin

91�� will soon offer a new doctoral degree in cybersecurity, a milestone reflecting both the rapid evolution of the cybersecurity field and the continued growth of the university's Department of Cyber Security Engineering (CYSE). The State Council of Higher Education for Virginia (SCHEV) approved the new PhD program in July.

"This is not just an administrative accomplishment," said CYSE Chair Paulo Cesar Costa. "It is an important recognition of the strength, maturity, and future direction of our department. It marks a major step forward in our ability to educate students in cybersecurity with the engineering depth and systems perspective that define CYSE."

He said, “The main challenge we faced with SCHEV was that we had to explain how cybersecurity was already a discipline of its own.” He points to the history of computer science as a useful comparison. Decades ago, computer science existed within electrical engineering before technological advances created the need for dedicated departments and schools. "We were able to convince SCHEV that the new discipline is actually overdue already."

As society becomes increasingly dependent on internet-connected devices, automated systems, and artificial intelligence (AI), the field is expanding beyond defending computer networks. "Cybersecurity includes a lot more than just hacking computers. There's the human factor, there's policymaking, there's how we train people. And now with the advent of AI, it basically turned around all the knowledge that we had about cybersecurity."

Cyberattacks that once required months of planning by highly skilled teams can now be executed far more quickly using AI agents, while researchers are already seeing examples of AI systems exploiting vulnerabilities in other AI systems.

"Cybersecurity and AI are no longer adjacent problems,” said Amarda Shehu, vice president and chief AI officer for the university. “AI supplies new defenses and new attack surfaces, and that demands researchers trained in engineering depth rather than in tools that expire. George Mason has been at the forefront of cybersecurity engineering for years, and this doctorate is the next step—producing those researchers at a public university whose obligation is to widen access to such expertise."

Costa said, "The whole discipline is evolving so fast that if we don't focus on this and get prepared, it's not that the CYSE department is not going to succeed. It's that the country will not have the capable cybersecurity professionals it needs."

For Costa, the PhD program is the latest step in a journey that began more than a decade ago. "When the cybersecurity department was created, it was created in a very unconventional way," he said. "It was created as a result of our own success in the bachelor's program."

The bachelor's degree launched in 2015 as a standalone academic program, with courses taught by faculty from other departments; enrollment quickly exceeded expectations. "It exploded in success," Costa said. "When it crossed 400 students, the dean's office created a new department."

The department later added a master's program, but Costa knew that “any department seeking to be known for its excellence has to have its own PhD program."

By 2023, he believed the department had the necessary infrastructure. "When I became chair, I allocated a lot of resources to start that program; it requires a lot more than just one guy writing stuff," Costa said.

Costa credited CYSE Professor Zhengdao Wang and Assistant Professor Tanvir Arafin for their extensive work developing and refining the proposal.

Wang said the program is designed to educate the next generation of cybersecurity experts. "We have an interdisciplinary curriculum that emphasizes security by design, integrates AI throughout the curriculum and research, and prepares graduates to address emerging challenges. The program will translate fundamental discoveries into practical solutions while strengthening the cybersecurity workforce and research enterprise.”

The college will implement a recruitment plan for the program soon, and it will be listed on the University Catalog in the near future.

Topics

Cybersecurity

Cybersecurity Research

Department of Cyber Security Engineering

Campus News

doctoral students

Digging out from disaster by digging into data

Nathan Kahl — Mon, 10 Aug 2026 16:52:30 +0000

Digging out from disaster by digging into data Nathan Kahl Mon, 08/10/2026 - 12:52

In This Story

Hemant Purohit

In the aftermath of a devastating earthquake, the world turns its attention to the affected region, watching search-and-rescue teams race to save lives, supplies pour in, and journalists talk to impacted residents. While the focus usually only lasts for several days, for those at the Humanitarian Informatics Lab in 91��'s Department of Information Sciences and Technology, another phase begins after the cameras leave.

"The response is critical, but recovery continues for months and years," said ab director Hemant Purohit, co-director of the Center for Resilient, Adaptable, and Safe Communities (C-RASC). That's where he feels he and his colleagues can make the greatest contribution.

Purohit’s approach builds on decades-long collective expertise of C-RASC colleagues and nearly a decade of work on human-centered AI technologies by the Humanitarian Informatics Lab, which develops technologies that transform large volumes of data into practical decision-making tools for disaster response and humanitarian operations.

A man and dog search through earthquake rubble. iStock photo.

For example, within days of the June 2026 earthquake in Venezuela, Purohit and colleagues at C-RASC began asking how George Mason could help organizations that will be rebuilding communities long after the immediate crisis passes. The answer became the Venezuela Earthquake Recovery Initiative, a collaboration between George Mason's C-RASC and the Gulfstream Foundation that is developing artificial intelligence and data-driven tools to support governments, nonprofit organizations, and humanitarian agencies during the long-term recovery effort. The team will soon approach the U.S. State Department about using their tools to assist the South American country as it recovers.

Lin Wells, executive advisor to C-RASC, noted that George Mason has much to offer. “From our years of experience, we’ve found that those who want to help must recognize technology alone is never enough. They must build trust with local people and develop solutions along with them. Sociology always tops technology. Also, rebuilding efforts should empower the damaged communities to keep developing on their own after the aid ends. Don’t just bounce back, bounce forward better.”

“Information is essential as interventions remain dynamic and earthquake visible effects are static,” said Gilberto Guevara, a humanitarian expert affiliated with GulfStream and C-RASC, with over 35 years of work in the humanitarian world. “One hopeful sign is to track progress and share the joy of constant activity. Information flows, validated and verified, are what make the difference and keep the momentum going.”

Rather joining the crowded emergency response underway, the team chose to focus where its expertise could have the greatest impact. "Showing up without the right capabilities during the response phase can be counterproductive," Purohit said. "Instead, we asked how research and technology could provide information for the organizations leading recovery."

One technology is Citizen-Helper, a platform launched in 2016. Originally developed through National Science Foundation–supported research, it uses social media and citizen-generated information to identify needs emerging in disaster-affected regions. During the Venezuela recovery effort, Citizen-Helper is being adapted to understand what people are requesting, where those requests originate, and how needs vary across different communities.

"If people are talking about shelter, clothing, medical care, or volunteers, we want to be able to identify those patterns quickly," Purohit said. "That kind of citizen-generated information can provide valuable information."

The team is also developing the AI-powered Recovery Intelligence and Support Engine (RISE), combining geospatial mapping with conversational artificial intelligence. The platform solves disaster recovery's persistent challenge of having too much information spread across too many sources.

"Our goal is to make complex information easier to access for the people making important decisions," Purohit said. "Instead of searching through scattered datasets, they can interact with the vast amount of information in a much more intuitive way."

The project grew out of ongoing collaboration with the Gulfstream Foundation, an organization with extensive experience supporting resilience efforts in Puerto Rico. During a regularly scheduled meeting the week of the earthquake, participants shifted from discussing existing projects to exploring how they could support recovery in Venezuela.

The team brings experience from multiple major disasters, including crisis mapping efforts, Puerto Rico recovery initiatives after the 2017 hurricane season, and humanitarian relief coordination since the Haiti earthquake in 2010. The informed strategy focuses on practical tools rather than theoretical research.

Future adaptation of these tools could help convert damage assessment data into recommendations for health care delivery, support temporary housing decisions, and provide analytics that identify communities at risk of being overlooked.

For Purohit, the initiative reflects the broader mission of the Humanitarian Informatics Lab and C-RASC: using engineering and artificial intelligence to strengthen the work of those serving communities during times of crisis.

Topics

Humanitarian Assistance

big data analytics

National Institutes of Health Javits Award to bolster building the brain’s blueprint

Nathan Kahl — Thu, 23 Jul 2026 18:20:03 +0000

National Institutes of Health Javits Award to bolster building the brain’s blueprint Nathan Kahl Thu, 07/23/2026 - 14:20

The goal of Giorgio Ascoli’s three decades of research is to gain a more complete understanding of one of science's most complex systems: the human brain. And this year the 91�� professor of bioengineering and neuroscience received one of the National Institutes of Health's (NIH) most prestigious recognitions for established investigators.

Giorgio Ascoli. Photo by Ron Aira/Office of University Branding

Ascoli's long-running program was selected for an NIH Javits Neuroscience Investigator Award, providing seven years of support and additional funding to advance his research into the structure and function of the brain.

The award is especially significant for George Mason, as Ascoli's is the first Javits Award recipient at the university. The award is reserved for a small number of exceptionally highly rated research proposals; the National Institute of Neurological Disorders and Stroke, the institute that supports Ascoli's research, only gives this distinction to a handful of grants each year.

Ascoli's proposal received a priority score in the top 3 percent of all applications reviewed. The Javits recognition converts Ascoli’s funding from what is categorized at NIH as an R01 grant—and which he has successfully renewed since 1999—into an R37, giving a longer timeline and more security. “Because there is a more extended time frame, it really gives the lab a chance to think longer term. I can now take research risks that I could not normally take.”

His work focuses on the intricate structures that allow nerve cells to communicate with one another. Unlike most cells in the body, neurons possess elaborate branching structures that extend from the cell body. These structures receive and transmit signals, creating vast networks through which the brain processes information.

Ascoli compares the structures to a blueprint. "If you're an engineer and you're given a machine and you want to understand how this machine works, the best thing that you can get is a blueprint," he said.

For decades, Ascoli's lab has charted the diversity and organization of these neuronal structures, creating quantitative models of the brain's cellular architecture. A central outcome of that work is NeuroMorpho.Org, a publicly accessible database that collects and standardizes data on the structure of neurons from research groups around the world.

“The impact of our research goes beyond the typical impact of a single lab, which is publications, discoveries, etc. This reaches out more broadly and with longer-term consequences for scientific research,” he said. Researchers worldwide contribute data, which are standardized, annotated, and made freely available for anyone to download, analyze, and use in computational models.

The Javits-supported research will build on that foundation by combining neuroinformatics, computational neuroscience, and the study of learning and memory. Ascoli and his team will use data-driven computational models to simulate brain networks and investigate how their structure supports the brain's ability to learn and retain information.

A particular focus is the hippocampus, a region of the brain closely associated with learning and memory. The work aims to better understand how the architecture of neural circuits contributes to different forms of memory, including conscious memories of events and experiences.

The research could help scientists make predictions about how the brain functions and then test those predictions experimentally. The approach also reflects the increasingly interdisciplinary nature of modern neuroscience, bringing together biology, computer science, engineering, and data science.

“Never before has the neuroscience community had access to so much data and such mighty computational resources for analysis and modeling,” he said. “We are tantalizingly close to making exciting breakthroughs in the century-old quest to link neural architecture to cognitive function.”

The Javits Award will give Ascoli and his team additional time and resources to continue building his brain blueprint, one neuron and one connection at a time.

In This Story

Giorgio Ascoli

Explore Research at George Mason

Topics

Neuroscience

Bioengineering

National Institutes of Health

NSF CAREER award allows researcher to look under AI’s hood

Nathan Kahl — Mon, 20 Jul 2026 16:30:48 +0000

NSF CAREER award allows researcher to look under AI’s hood Nathan Kahl Mon, 07/20/2026 - 12:30

In This Story

Ziyu Yao

Artificial intelligence (AI) systems are now part of everyday life, writing software, tidying up your shopping list, and answering questions. While we know they sit on top of large language models (LLMs), we still understand surprisingly little about what happens inside these systems or what is helping them generate responses.

Ziyu Yao. Photo provided

91�� Department of Computer Science Assistant Professor Ziyu Yao received a prestigious National Science Foundation (NSF) CAREER award for $674,100 over five years to peel the layers back on AI. Specifically, Yao will develop new methods for understanding the internal mechanisms of LLMs. Her research aims to make these AI systems more transparent and useful.

"We use language model–powered products every day, but we have very limited understanding of how they actually work," Yao said. Most efforts to evaluate AI systems focus on their behavior. Researchers give a model a prompt and judge the quality of its response. Yao's work instead focuses on mechanistic interpretability, an emerging field that examines what happens inside a language model as it processes information.

"If we send something to the model, we want to see how it extracts features or signals from my input and gradually performs computations before producing an answer," she said.

As language models have grown dramatically in size, that task is increasingly difficult. Today's leading models contain potentially hundreds of billions of parameters. And Yao said, "The interpretability research community has not caught up. Many traditional approaches cannot be efficiently applied to today's language models."

The project focuses specifically on AI-powered code generation, in which users describe a programming task in natural language and an AI system generates software code. Code generation is an increasingly important application of LLMs, with uses including software engineering and scientific research.

Yao's team will investigate how LLMs apply programming syntax, recall semantic knowledge, and perform deliberate reasoning while writing code. The researchers will also study how training data and learning methods influence those internal mechanisms and explore how interpretability insights can be used to improve AI coding systems.

Current interpretability studies often require researchers to manually design experiments, collecting internal model signals and analyzing the results. Yao plans to develop a framework that automates much of that process using AI itself, allowing researchers to study LLMs more efficiently.

At the same time, she wants interpretability research to produce practical benefits rather than simply satisfying scientific curiosity.

"We want to build the bridge between what we discover about language models and how that knowledge can improve them," Yao said. "Can interpretability actually help us build better AI systems?"

Although commercial AI systems such as ChatGPT are proprietary, Yao and colleagues primarily study open-weight language models that are downloadable. These models are becoming increasingly competitive with commercial systems while supporting the transparency needed for scientific research.

An expectation with the award and a hallmark of the CAREER program is a plan for broader impact. Yao hopes to expand AI literacy through outreach programs, create new curricula and open-source educational resources, provide undergraduate research opportunities, and foster collaborations between researchers in artificial intelligence and software engineering.

"I think we should encourage the younger generation to embrace AI," Yao said. "But at the same time, they should be skeptical, think critically, and be careful about what they are reading from AI."

Topics

large language models

National Science Foundation