Every year, Lawrence Technological University (LTU) recognizes an outstanding researcher from each of its five colleges. The College of Arts and Sciences (CoAS) selected Tao Liu, PhD, as its 2026 Researcher of the Year.
Interdisciplinary Expertise
Liu’s selection reflects his growing impact on research and program development at LTU. CoAS Dean Patrick Nelson, PhD, said, “Tao’s interdisciplinary expertise has strengthened connections between the College of Arts and Sciences and the College of Engineering. He has played a central role in building LTU’s growing Cybersecurity program—one that thoughtfully integrates software, hardware, and their intersection—making it increasingly popular among students.
“He is also a recognized expert in AI, a fact reflected prominently throughout the external letters and his scholarly record that includes 12 conference proceedings, three peer-reviewed journal publications, and five grant proposals submitted or in progress as PI or Co-PI. He has accumulated more than 1,090 citations and is approaching a position among LTU’s top 20 most-cited faculty—an exceptional achievement in such a short time.”
AI Expertise
Liu’s research lies at the intersection of algorithms, architectures, and security in artificial intelligence. His work focuses on the design, acceleration, and protection of AI systems, with particular emphasis on efficient, secure, and robust deep learning for real-time, embedded, and edge applications.
At Lawrence Tech, he leads research efforts on embedded deep learning, cybersecurity, and cross-domain AI. For example, Liu is exploring how emerging hardware technologies can reduce AI latency and power consumption.
AI latency simply means how fast an AI system can run and produce an answer after it receives information. For example, when you ask a voice assistant a question, the short pause before it answers is its latency. In everyday situations this delay may be small, but in applications like self-driving cars, drones, or medical monitoring systems, AI must respond almost instantly to ensure safety. Reducing this delay is challenging because modern AI systems often need to process large amounts of data and perform many calculations before making a decision.
Power consumption refers to how much electricity an AI system needs to run. Some powerful AI systems require many computers working together and can use a large amount of energy. One challenge researchers face is figuring out how to make AI systems that are still very powerful but use much less energy, so they can run on smaller devices such as phones, smart sensors, or wearable health monitors.
He also researches how AI enables new capabilities in autonomous systems, IoT devices, and cyber-physical systems. IoT or Internet of Things devices are physical objects embedded with sensors and software that connect to the internet to collect and exchange data. Cyber-Physical Systems integrate computation, networking, and physical processes, where IoT plays a crucial role in bridging the digital and physical worlds. For example, a fitness tracker on your wrist measures things like heart rate, movement, and sleep patterns. With AI, the device can learn from this data to better understand a person’s body. Instead of simply counting steps, the AI can analyze patterns and provide smarter insights, such as detecting unusual heart rate changes, estimating stress levels, or suggesting when someone should rest or exercise. This helps turn simple sensor data into more accurate and useful information about a person’s health.
A key component of his work involves adversarial machine learning, where he investigates vulnerabilities in AI models and develops defenses to ensure safety, reliability, and trustworthiness in mission-critical domains. In simple terms, this research studies how someone might try to trick an AI system—and how to stop that from happening. For example, a tiny change to a picture might cause an AI system to mistake a stop sign for something else, even though it still looks normal to people. By studying these weaknesses, researchers can make AI systems stronger, safer, and harder to fool.
By: Renée Ahee
